Traversing the Labyrinth: A Comprehensive Analysis of Pedestrian Traffic in Venice An Interactive Qualifying Project report submitted to the faculty of WORCESTER POLYTECHNIC INSTITUTE in partial fulfillment of the requirements for the Degree of Bachelor of Science.
Submitted on December 15, 2011 by: Chelsea Fogarty Geordie Folinas Steven Greco Cassandra Stacy
Project Advisors: Professor Fabio Carrera, Ph.D. Professor Frederick Bianchi, D.A.
Project Information: ve11-mobi@wpi.edu https://sites.google.com/site/ve11mobi
Sponsors: Venice Project Center Venice Department of Mobility
In Collaboration With: Santa Fe Complex Redfish Group
Abstract The 2011 Mobility Interactive Qualifying Project focused on pedestrian movement in the San Marco district of Venice, Italy. Congestion often occurs in the city’s streets, and has a negative effect on quality of life. To better understand this problem, the team quantified, analyzed, and publicized pedestrian traffic at 7 bridges, 4 traghetto stops, and 3 Actv stops within the study area. Data was also compiled from census tracts, the Actv, and other sources. Utilizing the collected data, a computer model of the dynamic movement of pedestrians was created by the team’s collaborators. This computer model aptly modeled traffic within the San Marco district and can be used as a proof of concept for a city-wide model that would be able to forecast instances of high pedestrian volume. Forecasting pedestrian congestion would allow preventative measures to be employed and improve traffic flow throughout the city.
1
Acknowledgements The 2011 Mobility team would like to extend their deepest thanks to their sponsor at the Venice Department of Mobility and Transportation, Signore Loris Sartori. Without him, the project would not have been as successful or comprehensive. They would also like to thank the Venice project advisors, Professors Fabio Carrera and Frederick Bianchi, for their valuable input and guidance. Lastly, they would like to thank Cody Smith from the Santa Fe Complex, who provided indispensible coding knowledge in respects to the agent-based model.
2
Authorship This Interactive Qualifying Project was completed with contributions from each team member.
3
Executive Summary Venice, Italy is a city composed of islands that are separated by canals and connected by bridges. It can only be traversed by means of boat travel or on foot. These two modes of transportation are greatly independent, but coincide at public transportation boat stops. The main focus of this project was pedestrian traffic, on which data was quantified using an efficient field counting methodology, analyzed, and distributed to a third party collaborator to include in an agent-based model. Pedestrian traffic is often impeded by narrow walkways, bridges, street vendors, and public tourist attractions. These factors can impede movement and contribute to congestion. The maximum carrying capacity reported for Venice has been determined to be approximately 25,000 tourists per day, which is frequently surpassed and is the main contributor to the issue of pedestrian traffic congestion in the city1. The native population is approximately 61 thousand people, and the amount of tourists on many occasions outnumbers the local population by more than double. Consequently, many residents relocate to the mainland to escape the high volumes of tourists. The overwhelming amount of pedestrians that flow in and out of the city creates a need for a better understanding of pedestrian traffic flow and efforts towards improving mobility efficiency within the city’s infrastructure. This will reduce the pressure on the city itself and on the occupants, and ensure easier transportation in and out and within the city. The mission of this project was to collect pedestrian traffic data for the end goal of developing a modeling system that collects and archives data to effectively predict the behavior of pedestrian mobility in the San Marco sector of Venice. This model will improve mobility efficiency throughout the sector. We worked to accomplish this goal by quantifying pedestrian traffic at critical flow points and integrating this data into an autonomous agent computer model to visualize mobility. Additionally, we developed a comprehensive methodology and procedure for employment by future IQP groups and other groups who wish to improve pedestrian mobility efficiency.
1
 Swarbrooke, John. Sustainable Tourism Management. N.p.: CABI, 1999. 4
PEDESTRIAN TRAFFIC STUDIES To analyze pedestrian traffic, this team focused on collecting data on different aspects of mobility within the district of San Marco. Focusing on this area allowed for a thorough study of pedestrian mobility and the development of a resourceful visualization of the traffic in the computer model. Population was categorized into two agent types—Venetians and tourist. Each type was quantified at seven major bridges and matched to their likely destinations and origins throughout the day using census data and data on known attractor locations. Several days of preliminary field counts determined high volume time brackets as well as assumptions that can be implemented while collecting data. Counts collected using video camera feeds recorded during several different pedestrian traffic scenarios and at various angles we used to determine the feasibility of collecting data from video surveillance cameras. Ridership data available from the public transit system (ACTV) as well as pedestrian counts performed at four traghetti stops were analyzed to determine boat usage. At the three high-volume bridges at which data was collected, the following maximum traffic levels by proportions of Venetians and tourists were determined:
5
San Moisè experienced the most pedestrian usage; however, Venetians and pedestrians almost equally use it at the time when it reaches maximum volume, while the other two bridges are dominated by Venetians. The following map displays the traghetti usage throughout the day by arrival and departure into and out of the San Marco district:
6
Sant’Angelo is a more frequently used stop, mostly because it operates for more hours than the other two traghetti stops. Also, both Carbon and Sant’Angelo are used to enter San Marco district, while San Samuele is used mostly to Exit San Marco. This could be contributed to the fact that San Marco is densely populated with residencies and the work locations are outside of the district. So everyone that lives near the San Samuele stop is using it to cross over to his office location. Sant’Angelo is not significantly used to either exit or enter the district. Future studies can supplement the data collected with additional data from the remaining bridges and traghetti in the San Marco district, as well as with data from other districts to visualize mobility throughout the entire city of Venice. Traffic levels throughout the year, especially during the tourist season, as well as the effects of weather conditions should also be studied. Video surveillance installed at bird’s eye angles at each major bridge location and traghetti stop, and the video surveillance already installed at ACTV stops is recommended to collect counts year-round for daylight hours. 7
AUTONOMOUS AGENT COMPUTER MODEL
CONCLUSIONS
8
Introduction Cities worldwide adopt a negative reputation for their mobility issues. Mobility is the freedom to move about, and when mobility is impeded, people are forced to interrupt their routes and pace and accommodate for their lost time. Many travelers try to avoid city traffic to save time on their trip, and those who cannot avoid traveling through cities must plan ahead accordingly. In an attempt to better increase mobility, urban districts adopted public transit systems in the form of buses, underground subways, trams, trains, and even boats. These systems can transport large amounts of travelers and ease the congestion that results from high usage of private transportation. Mobility issues are even more perceptible in Venice, Italy because the only modes of transportation are by foot through complicated walkways and over narrow bridges, and by boat.
Figure 1 A map of the framework of Venice's islets, canals and walkways.
The city is made up of 121 islets connected by 435 bridges2, with no room to expand. The branch canals range from 10 to 30 feet in width, and the intricate network of walkways are made up of streets of no more than seven feet wide; the widest don’t exceed twenty feet3. In 2008, the City Council Tourist Department released its annual report, which stated that 5,875,370 people visited
2http://www.comune.venezia.it/flex/cm/pages/ServeBLOB.php/L/EN/IDPagina/117 3(Morgan
1782)
9
historic Venice4 in 2007. 16% were Italians, and the rest were foreigners. This figure has doubled since the 1980s. Case studies run in 1991 determined that the carrying capacity of Venice was 25,000 visitors per day5. As illustrated in Figure ##, tourism is consistently increasing in Venice, but the infrastructure is limited to the amount of pedestrians it can containcannot expand, and so is at risk of succumbing to the mass amount of pedestrians traversing the city on a daily basis.
Tourist Forecast within Historic Districts 40,000,000 35,000,000 30,000,000 25,000,000
15,000,000
Overnight Tourists Day Trips
10,000,000
Total Tourists
5,000,000
Expon. (Total Tourists)
20,000,000
0 1940
1960
1980
2000
2020
2040
2060
Locations that often create holdups in traffic are called bottlenecks. Bridges are evident locations where traffic jams frequently occur, especially when tourists stop at the top to take pictures of the view. Alongside Other than bridges, pedestrians in limited amounts can get from island to island using the gondola di traghetti or the Azienda del Consorzio Trasporti Veneziano (ACTVActv), the public boat transportation system. These forms of boat transport have helped alleviate a portion of the overcrowding at bridges as well as facilitate the flow of water traffic by centralizing water travel
4http://www.aguideinvenice.com/en/venice-case-8-Report-on-tourism-in-Venice-December-2008.html 5
Swarbrooke, John. Sustainable Tourism Management. N.p.: CABI, 1999.
10
Comment [CS1]: This doesn’t matter as much as the quantity of people. Comment [CS2]: The number of tourists or the % of Italians? Comment [CS3]: Relevance? If we go by the 2007 number of tourists, it’s only 16,000 per day, which doesn’t exceed the carrying capacity.
through 20 routes on the canals, as seen in Figure ##. However, at certain times of the day, waiting for space on these two types of boat transportation slows pedestrian movement down.
Figure 2 A map of the ACTV public transportation system.
On various occasions, such as the Carnival and other festivals, trafficcongestion can become so severe that pedestrians come to a standstill. During these instances, the city must take reactive measures to alleviate congested areas. These reactive measures include calling upondeploying the police last-minute to go on site and direct traffic flow, or temporarily making walkways unidirectional. However, by the time officials get to their stations the traffic is already at a severe state, and when pedestrians are not informed of changes in accessibility, their routes must be amended unexpectedly. If there were a preventive measure installed so that city officials could predict traffic behavior, then traffic can could be dealt with before it reaches an extreme state level and events and transport can operate more smoothly.
11
Comment [CS4]: Explain this statement.
Figure 3 At some instances, streets can become so crowded that traffic is at a standstill
For the past several years, the Worcester Polytechnic Institute Venice Mobility Interactive Qualifying Project teams have been working with the Department of Transportation and Mobility, collecting qualitative pedestrian data with the intention of creating a computer model to be used as the method fordeveloping different means of preventing traffic issues. There have been some holes in the data and execution, however. Individuals are responsible for collecting data on-site, risking the
Comment [CS5]: Just because the model is really only this year. Last year was a mobile app, so we can’t specify the model.
chance for human observational error. Data is collected in intervals of time and only in the tourist
Comment [CS6]: What does that even mean??
off-season when Worcester Polytechnic Institute Interdisciplinary Qualifying ProjectWPI IQP groups are on location. Also, the data that has been collected in the past disappears with time because there is no central database for archiving data. In order to create a comprehensive pedestrian computer model, there should be an automated data collection method so that data is continuously collected and archived in a public online resource. The city has several observational systems installed that would be advantageous for the purpose of preventing traffic issues. These surveillance systems are the Automatic and Remote Grand Canal Observation System (ARGOS), Hydra, and Security and Facility Expertise (SaFE) and they are placed in strategic locations throughout Venice that give them the ability to allow data to be collected not necessarily in real-time, but off of video clips that can be played back. Currently, these observational systems are used to implement speed limit laws, and monitor pedestrians and boats for crime. ARGOS gives the vigili urbani (the Venetian police) the opportunity to routinely dispatch officers to control traffic and make arrests on the Grand Canal, and Hydra and SaFE allow authorities to monitor the Venetian ports for potential crime6. If these cameras, as well as other 6(Bloisi,
et al. 2009)
12
cameras that could be installed in the future at other tactical locations, were used to collect traffic data, the data could be collected at all times of day and all year round. Clips could also be rewound and slowed down, to make sure that observational counts were collected as accurately as possible.
Figure 4 With the ARGOS system, live images are stitched together to generate a view of the Grand Canal. Observations are used from a multi-step Kalman filter to track targets over time7
In an attempt to combat the pedestrian traffic congestion that plagues Venice, there must be an efficient method to understand it. There is an abundance of information available from the ARGOS, Hydra, and SaFE systems, as well as research from WPI IQPs, but there is no tool that combines all of the data and presents it in a way that can be easily used in traffic prevention. This year’s Mobility team and collaborators created a model that will eventually accomplish this throughout the entirety of Venice. In order to create a comprehensive pedestrian computer model, there should be an automated data collection method so that data is continuously collected and archived in a public online resource. Another key advantage that video surveillance has is that it can be paired with computer software that distinguishes between different types of pedestrians, which are referred to as agents for the purpose of the computer model. The benefits of this identification feature in data collection is that each agent type will have its own behavior and walking speed and will go to different points of attraction. In Venice, pedestrians can be broken down into two simple agent categories: Venetians and tourists. Tourists have a random walking pattern, being attracted to museums, churches, and shopping centers, while Venetians have a more structured pattern to and from home or work. In order for the model to accurately predict the flow of traffic, it must be able to illustrate the differences in walking patterns between locals and tourists.
7
http://www.dis.uniroma1.it/~bloisi/segmentation/segmentation.html#ARGOS_Project
13
Establishing a framework for the collection of data and developing the database for the computer model is where the 2011 Mobility team came into play. A structured methodology for collecting and archiving data has been instituted and can be executed by future traffic improvement teams. Additionally, this methodology has been executed at key bottleneck locations throughout the San Marco district for the continued development of this system. This data was integrated into the beginnings prototype of the agent-based computer model designed by the Mobility team’s collaborators, along with data compiled from various sources provided by the municipality of Venice, that is to be continuously added upon. The end goal of this project iswas to have a pedestrian agent-based automated model that will predict the flow of traffic efficiently for the benefit of alleviating traffic congestion throughout the streets of Venice.
14
Background Venice is composed of canals and narrow streets, which makes it a one-of-a-kind city to travel through. Though the historic city occupies merely three square miles of land, traveling quickly and efficiently can be a challenge due to a complex web-work of walkways, overcrowding in areas and events that attract tourists, a disconcerting water bus schedule with many different routes and times, and severe weather conditions where flooding can occur, forcing pedestrians to have to walk on platforms, further narrowing the plane of mobility. For the uninformed, moving through Venice can be an unnecessary crusade.
2.1 THE ARCHITECTURAL FRAMEWORK OF VENICE Venice is a very small, yet multifaceted city that has changed its role many times over the years. Now known as a “museum city”, it was not originally meant to be an attraction for people all over the world. The city was not meant supposed to hold as many people as it frequently does. Because of Venice’s physical limitations, it has a difficult time accommodating for the congestion issues that result from the mass influx of tourist. 2.1.1 Origins of the City
Venice is a city frozen in time. Its peculiar situation and magnificent architecture render it unique and peerless even in its decadence. How a city can be afloat in the sea and still be habitable and beautiful is marvelous. Interestingly enough, Venice originated in an “expedient of desperation” and became great by “accident of position8.” The city began as a collection of inhospitable islands in the Venetian lagoon, along the western shore of the Adriatic Sea. The invasions of the Lombards into northern Italy in AD 568 drove many mainland Italians onto a group of islands of the lagoon, which were originally the homes of traveling fisherman and salt workers9. Because the canals and rivers were not easy to navigate and the lands were unwelcoming, the islands provided excellent protection against possible naval attack. The population of the new Venice revolutionized the balance of forces throughout Italy. All facets of society from the mainland were preserved along with their various rights and social roles. Among them were the leading members of their ecclesiastical hierarchy.
8(Morgan 9(Cessi,
1782) Cosgrove and Foot, Italy 2011)
15
Comment [FB7]: There is a problem with the organization of the paper…you start this section with a discussion of Venice and it’s geography etc…. then you lead into a statement about the problems with congestion….then…you go back to describing Venice and its lagoons, etc….I can’t follow the development of ideas in the paper….think about the organization…..
Waves of refugees continued to flow onto the islands as the Lombards gradually took more territory from the Byzantines until AD 639 when the fall of Oderzo solidified the collapse of the Byzantine defense system10. This was a key moment in the emergence of maritime Venice. Venice was still loyal to the Byzantine government, and therefore all public administration was still carried out in its name, yet the continuing war against the Lombards eventually brought strain to the government’s control of the city. The pressure of wartime life increased the Venetian’s inclination towards independence. The outbreak of religious conflict between Rome and Byzantium around 726 created serious clashes in Italy. Venetian troops joined forces with the Pope and took a stand against the authority of the exarch, electing the first doge, Duke Orso, while still remaining under the Byzantine Empire. It was not until the collapse of the empire in 751 that independence was accelerated. While Venice was dealing with political strife and continuous turnovers of power, it was also taking advantage of the opportunities offered by the sea and commerce. Trades passing through the city included dyes, leathers, spices, and many other goods. The lagoon province was the bridge between the European west, and the Islamic and Byzantine territories in the east. By the first half of the sixteenth century, Venice was the “great metropolis” that it is well-known for today. It hosted a variety of activities, trade continued on a large scale, and people came from all over the world. 2.1.2 Design of the City
What once was a group of islands with wooden houses resting on poles staked into unstable clay soil gradually morphed into an elegant and romantic city of stone. The buildings had to be strategically placed, taking into account the special environmental conditions of Venice. Weight had to be properly distributed so that there were never too many areas of stress11. Population and manufactures grew exponentially, and because the city could not expand outward, it expanded up. It was also less expensive to build another floor than to buy more land. Buildings were built close together, and very tall. The ground floor usually housed businesses, while the upper floors provided homes for families. As the city grew and its economy became prosperous, the structures reflected the transformation. The principal buildings in Venice were constructed of marble or light stone, and the remaining were 10
(Ortalli 1999)
11(How
Were Houses in Ancient Venice Designed and Why? n.d.)
16
of brick covered with mastic for adhesion12. Palaces and houses were built and rebuilt overtime, along with churches, monasteries, and bell towers. The shape and direction of the canals were changed and bridges, road systems and boat transportation were integrated. Various architectural styles such as the famous Gothic, Roman, Byzantine and Renaissance techniques were blended together. The architecture and design possesses characteristics of permanence and timelessness that is unsurpassable. 2.1.3 The Canals
The employment of a network of canals in place of streets was more a matter of necessity than of choice. The current canals circumscribe the original islands, while the rest of the water area has been recovered by erecting walls composed of granite along the line of these canals, which lay the foundation for the adjacent buildings13.
Figure 5: A Canal Near the Arsenale
The branch canals off of the Grand Canal are some fifteen feet wide, and are often crooked and short in length. The Grand Canal is one of the major water transportation corridors in the city; it stretches down the center of the city in a backwards S-shaped course and is approximately 2 miles in length, 30 to 70 meters wide14. The sides are lined with palaces and buildings reflecting the Gothic, Romanesque, and Renaissance grandeur from its early development.
12(How
Were Houses in Ancient Venice Designed and Why? n.d.) 1782) 14(Cessi, Cosgrove and Foot, Italy 2011) 13(Morgan
17
2.1.4 The Streets
There are 2,194 streets, each one as unique as its canals, which make up the labyrinth that is the city of Venice15. They too are narrow, short, and crooked, and they penetrate every part of the city. Most of them are passages about seven feet wide, with the widest of streets not more than twenty-five feet16. Some terminate abruptly and turn at sharp angles. Every street is covered with pavement, and on each side are gutter stones to pass surface water or rain into conduits underneath17. While the picture of these streets sounds uninviting, the close proximity is relieved by the numerous squares that intersect them. There are 294 squares scattered throughout the city18. The streets cross the canals by means of 409 bridges, consisting of a single arch, with a roadway graded into low steps, connecting every bit of land in Venice19.
Figure 6: A Standard Street in Venice
2.2 ENVIRONMENTAL IMPACTS ON MOBILITY Venice’s unique infrastructure is slowly degrading from the severity of the environmental impacts it sustains. The city’s environment is “… suffering from a general hydrogeological imbalance which is dramatically evident in the erosion of the lagoon morphology and in the number of exceptional high water events” in Venice20. This has been a problem for centuries, and the occurrence of tides high
15(Morgan
1782) 1782) 17(Morgan 1782) 18(Morgan 1782) 19(Morgan 1782) 20(Rameiri, et al. 1998) 16(Morgan
18
enough to flood, called acqua alta, has been increasing at an alarming rate: from four to five times per ten years at the turn of the 20th century to at least thirty times per ten years today21. 2.2.1 Acqua Alta
The phenomenon of acqua alta occurs when there are southeast winds and a high tide at the same time, which causes the waves to spill over the canal walls into the city streets22. When water overtakes the walkways, pedestrian traffic flow is slowed and the area in which pedestrians can travel is limited, creating severe congestion. Sidewalks become flooded when there is a tide 100 or more centimeters above the average sea level. Platforms raised 120 centimeters off of the ground, called passerelle, are placed strategically along flooded pathways to enable pedestrians to walk above the water. While this is a helpful and necessary strategy for staying dry, it has a severe impact on the walkers’ mobility. The passerelle are narrow and create a difficult passing situation. The cramped space makes the walking rate slow and creates pedestrian congestion. St. Mark’s Square, a popular tourist destination, is one of the lowest sections of the city, and as a result is flooded with every acqua alta. The passerelle are placed throughout the square and leading to other tourist destinations, and many tourists travel upon them. Since the platforms are “just barely wide enough for two-way traffic,” a tourist taking pictures or an older person walking slowly can cause a large section of the walkway to become congested23. If the tides rise higher than 120 centimeters, the passerelle are at risk of floating off of their supports. When this happens, walkways are completely hindered and only those with rainboots can walk through the city without wetting their feet. 2.2.2 Canal Wall Damage
Acqua alta is also a contributor to the erosion that is impacting the city so severely. The other large cause of erosion is the wakes caused by motor boats. As water collides with canal walls, it erodes the mortar that acts as an adhesive between the bricks and stone, and the wall becomes “more susceptible to the destructive stresses and forces” of the tides and wakes24. When the erosion becomes dangerous for pedestrians or the infrastructure, the walls must be repaired. Construction is necessary, but impairs mobility because the materials and space required for restoration overtake
21(Rameiri,
et al. 1998) (Davis and Marvin 2004) 23 (Davis and Marvin 2004) 24 (Black, et al. 2008) 22
19
parts of the walkways. This causes backups down the walkways and has an overall negative effect on congestion.
2.3 TOURISM IN VENICE The Queen of the Adriatic has been attracting foreigners for centuries, and according to Riganti and Nijkamp, the city can be considered a mature tourist destination: one that witnesses negative environmental impacts caused by tourist congestion more frequently than other destinations25. The magnitude of tourists that visit Venice has a huge negative impact on the city. The resulting congestion causes mobility impairments throughout the city, and especially at top tourist locations and during peak tourist times. 2.3.1 Popular Tourist Sites and Events
The concentration of tourists is a problem that Venetians have been attempting to control for a very long time. There are a number of specific locations throughout the city that are typically visited by tourists, which creates congestion both en route to the destination and at the attraction itself. The Piazza San Marco, or St. Mark’s square, is a popular tourist stop, where one can visit St. Mark’s Basilica and bell tower. Another is the Ponte di Rialto (Rialto Bridge), a large bridge connecting one side of the Grand Canal to the other with shops along it. These destinations, as well as many other spots in Venice, are the cause of the large amount of pedestrian traffic that regularly occurs. Beyond the draw of the city itself, there many events held in Venice that attract a high number of tourists annually. The Carnevale di Venezia, or Carnival of Venice, takes place in February every year, and marks the beginning of Lent. A huge amount of tourists travels to Venice to witness the Venetian beauty and culture displayed throughout the Carnevale and to attend the various events held during it, such as La Biennale (a contemporary art festival highlighting architecture, independent films, and paintings, among other things) and the Vogalonga (a boat race through the Venetian lagoon)26. Events such as the Carnevale lead to an extremely high tourist volume, which in turn causes mobility impediments for pedestrians attempting to travel from one place to another in an efficient manner.
25
(Riganti and Nijkamp 2008) di Venezia 2012 2009)
26(Carnevale
20
2.3.2 Magnitude of Tourists
The sheer magnitude of visitors to the city creates issues within the infrastructure and community. Traveling around world was once reserved for only the rich or influential, but it is now a viable experience for a majority of people. This evolution towards “mass tourism” is one that is clearly seen in Venice, where there has been a significant influx of tourists over the years27. The carrying capacity of Venice, or “the maximum number of visitors the attraction can handle at a given time without either damaging its physical structure or reducing the quality of the visitors’ experience” has been determined to be approximately 55,000 tourists per day28. This capacity is regularly surpassed, and that leads to the ultimate issue of Venetian traffic congestion. This congestion can be seen at tourist sites and on bridges, where the limited space often creates crowds of people trying to push through to their destination. Venice is becoming a European Disneyworld, or a museum city, where the tourists outnumber the natives: “[w]ith its thirteen million or more annual visitors and a local population of only around sixty-five thousand, historic Venice has the highest ratio of tourists to locals of any city in the world.”29 This overcrowding effect impairs and changes many aspects of life in Venice, not the least of which is commuting to and from work or attempting to traverse the city for another purpose. All of the factors described above: popular tourist spots, large events, and the city itself, cause an increase in tourists visiting Venice every year. The mobility impairment created by this group of people is severe, and must be addressed. The inability to traverse across the city lengthens work commutes for the employed and school commutes for students.
2.4 MOBILITY IN VENICE Due to its unique location, the city required extensive draining and dredging to provide more land to further the development of Venetian infrastructure. These operations led to the development of the first canals, and a rather unique system for the city’s mobility30. Transportation in the city exists in three main entities: the canals, bridges across them, and an arrangement of walkways. This network of more than 200 canals became a staple for the transport of goods throughout the city as well an excellent form of transportation. 27(Zanini,
Lando and Bellio 2008) (Venezia 2011) 29(Davis and Marvin 2004) 30 (Howard and Quill 2002) 28
21
2.4.1 Watercraft in Venice
Transportation and distribution of goods via the canal network would be impossible without the use of watercraft. Throughout history, all major cargo shipments and heavy transport is done by boat. For example, gondolas are iconic boats of Venice which were once used by the wealthy for transportation31. These boats are keel-less and used almost exclusively for tourism in this day and age32. Gondolas became far less popular with the development of steam powered vessels, called vaporetti, in 1881. These vessels are still the dominant form of nautical transportation in the city. Venetian ferries, called traghetti, are unglorified gondolas which are another popular form of transportation in Venice, and there are now seven of these ferry crossings across the Grand Canal33. These ferries operate at certain points between bridges on the Grand Canal and shuttle pedestrians across for just 50 cents34. Larger boats are used in Venice for cargo shipments, as well as for sea trade throughout the Mediterranean. Due to this demand for large ships, and a lacking of local resources, many Venetians became expert shipbuilders35. During the Medieval Era, Venice became one of the mightiest cities because of this drive for mercantilism. Venice was a major port along many trade routes which connected Europe to other continents such as Asia through the use of the Mediterranean Sea36. Venice also had a very well equipped navy, which had the ability to build one war galley per day37. These galleys were handcrafted in shipyards called squeri where all types of traditional boats were crafted. 2.4.2 Water-Based Public Transportation
Private boats are less common in Venice than watercraft used for shipping cargo and public transportation. This is largely due to the existence of taxi boats and a lack of space for extended docking. Taxis in Venice are multipurpose boats which not only transport clients to their desired destination but will also serve as a means of transportation for goods when not serving pedestrians. There are also other vessels which have scheduled routes throughout the city which can be used to move people between specified stops. (Cessi and Foot, Venice 2011) (Cessi and Foot, Venice 2011) 33 (Drake 2008) 34 (Drake 2008) 35 (Davis and Marvin 2004) 36 (Davis and Marvin 2004) 37 (Davis and Marvin 2004) 31 32
22
These forms of public transportation are one of the leading causes of boat traffic in Venice. Both taxis and gondolas have random travel routes, depending on their clients’ demands, and therefore become difficult to obtain data on. For example, gondolas typically serve as sightseeing vessels for tourists and will typically slow down and make stops near points of interests38. These stops can cause a large amount of traffic and affect mobility. The traffic patterns of taxis and gondolas are difficult to predict and their destinations are random, therefore their traffic patterns do not significantly influence overall mobility in Venice. 2.4.3 Pedestrian Mobility
The other prominent form of transportation in the City of Venice utilizes an array of walkways and bridges. The problems associated with these walkways are derived from how the city was constructed, which led to limited space, and an increasing number of tourists which visit the city. As the city was being constructed, walkways were built to facilitate trade and commerce in the city. Due to the significant space constrictions associated with construction on an archipelago, many buildings were constructed to the edge of the property, leaving little space for these additional walkways. This fact has left many of the walkways narrow, some spanning only about a meter across39. The stark narrowness of the walkways contributes to much of the pedestrian related traffic which occurs in the city, but it is not the only factor involved. The layout of the walkways has been compared to that of a labyrinth as a result of many canals being paved over to broaden the network of walkways and alleviate traffic demands40. Pedestrian traffic demands have been growing perpetually since the1950’s due to the overwhelming influx of tourists41. The combination of a large population of tourists new to the area and a confusing layout intensifies the effects of pedestrian congestion. 2.4.4 Venetian Bridges
The different islands of the archipelago are interconnected by an array of over four hundred bridges42. These bridges are crucial to the infrastructure of Venice, and have become recognizable as indispensable monuments of the city which are utilized on a daily basis43. Four of the most well-
(Chiu, Jagannath and Nodine 2002) (Davis and Marvin 2004) 40 (Davis and Marvin 2004) 41 (Van der Borg and Russo, Towards Sustainable Tourism in Venice 2001) 42 (Davis and Marvin 2004) 43 (Contesso 2011) 38 39
23
known bridges in Venice traverse the Grand Canal, including the Ponte di Rialto, Ponte dell’Accademia, Ponte degli Scalzi, and the most recent addition, the Ponte della Costituzione. The Ponte di Rialto was constructed in 1588, but initially had two predecessors. In 1175 a bridge was constructed using boats for floatation to span the canal, called a pontoon bridge, in the same location as the Ponte di Rialto44. This bridge was ultimately replaced in 1265 by a fixed bridge which later collapsed45. The Ponte di Rialto remained the only location to cross the Grand Canal until 185446. Today, pedestrians can cross the Grand Canal by using one of the four bridges which now exist, in addition to the seven different traghetti locations.
2.5 VENETIAN TRAFFIC MODELS Looking into future applications of data collection, the creation of an integrated pedestrian traffic model is necessary to provide an easy means of extracting useful information. Though the development of such a comprehensive model is out of reach for this year’s Mobility team given the time and fund limitations, it is important to understand pedestrian models so that data collection can be tailored to provide the models with information that is useful to its creation. The modeling approach that fits the needs of the Venice traffic model is referred to as agent-based modeling, and more specifically, autonomous agent-based modeling. This type of modeling allows for individual governing of agents, which lets each agent uniquely interact with the environment based on programmed predispositions and reactions. In modeling of traffic, each agent will be assigned a specific start and end location. Though the beginning and end are predefined, the method of transportation and the path taken vary based on the interactions between the agent and its surroundings, including other agents. In terms of Venice, agent-based modeling allows for the important distinction between tourists and locals in pedestrian mobility stream models. The accuracy of such a model is proportional to the agents’ ability to mimic the real life counterpart. Hence it is important to collect data that can speak to the various biases of agents. 2.5.1 Past Models
Since the beginning of the Venice Project Site, there have been several Interactive Qualifying Project teams that have done work that helped further traffic models. In 2008 a team created a pedestrian (Contesso 2011) (Contesso 2011) 46 (Contesso 2011) 44 45
24
Comment [C8]: Make sure to tie everything back to our project specifically. The past models don’t matter if they don’t apply to our project Comment [C9]: I don’t know. Weird sentence structure and is it necessary?
model using NetLogo, an agent based modeling environment47. The model focused on Campo San Filippo e Giacomo due to project time and resource constraints. This spot was chosen because it was identified as a hotspot, or high traffic area. The model accounted for Venetian and tourist agents and dictated their speed based upon data collected during the IQP. The model only portrayed traffic during Wednesday at 1300 hours due to data limitations. The data collected by the team during the IQP was inputted to the program. This data was collected and recorded visually using three cameras set up strategically around the hotspot48. Though the model created was limited and didn’t accurately portray congestion, it still demonstrates the necessity of an experienced programmer in creating a model, and demonstrates one accurate data collection technique. The importance of recording visual data should not be underestimated. It is crucial to confirming and checking past data collection.
Comment [C10]: Potentially opinion?
There was also a traffic model created in 2010 that detailed boat traffic in the city. This project was called Venice Table. The programming aspect was spearheaded by RedFish group and the Santa Fe Complex, with the Venice Mobility team providing the data for the model along with several government agencies. To allow for a comprehensive model of canal traffic, 23 observation points were used for data collection. In order to determine when each boat turns in the model, the data that was utilized included which canals boats entered from and returned to, the time of day, and each boat’s license plate number49. Control of the model was designed to be interactive and intuitive. To allow for the intuitive nature of the Venice Table, the model was built on an interactive gaming software program. 2.5.2 Modeling Tools
Traffic models are very useful tool for understanding and improving mobility streams. Unfortunately, the creation of good models takes time, expertise, and data. The implementation of an autonomous data collection system will allow the collection of data with minimal human interaction. There are several tools present that can make this type of continuous autonomous data collection a possibility. One of those tools is Open CV, which is a software approach that uses video to autonomously recognize, track, and record traffic and distinguish physical differences, as well as velocity.
Comment [C11]: Connect this to our project
(C. Catanese, et al. 2008) (C. Catanese, et al. 2008) 49 (VeniceTable: Interactive Traffic Simulation Table 2010) 47 48
25
2.5.3 How Models Read Data
Over the years, Venice has had countless groups, individuals, and governments study it and collect a wide array of data relevant to traffic. The question therefore becomes “How is this data formatted so that it can be inputted into a model?” The agent based models have proved useful in the past and will continue to be a method of data presentation. Agents, in our case pedestrians, will interact with the environment, Venice, developed in the model. The environment itself is made up of two main
Comment [C12]: Word choice?
components; edges and nodes. Edges are the borders and boundaries that define the fields in which the pedestrian agent types move. Nodes, on the other hand, are not physical or visible entities in the final 2D model. They help to define how the pedestrians will move. For instance a specific pedestrian, depending on the constraints that are programmed into a model, will move from a node ‘A’ to another node ‘B’. For the Venice models, these nodes are typically placed at traffic ‘choke points’ like bridges. For instance, a bridge spanning a canal in an east to west direction might have a node ‘A’ on its east side and another node ‘B’ on its west side. Movement defined as ‘AB’ would indicate a pedestrian moving from ‘A’ to ‘B,’ or one traveling west across the bridge. Movement defined as ‘BA’ would indicate the opposite: a pedestrian traveling east across the same bridge. Therefore data is organized by the number and type of pedestrian, as well as their node movement at choke points. Nodes can also help define sources (points where pedestrians originate) and sinks (points where pedestrians are attracted). How agent types are programmed will determine their ‘source-sink interaction’. In Venice, sources and sinks can be split up into two categories based on the types of pedestrians. Locals tend to originate from residential areas and will generally flow to places of employment or learning. In this case, this would mean that their homes are the sources and their places of work and schools are the sinks. At the end of the day, this would be reversed and the sources and sinks would switch. Tourists tend to originate from hotels, bus terminals, and the train station, and are attracted to places like museums, shops, and the “tourist triangle”. In the case of a museum, two nodes would still have to be used to define movement ‘in’ and ‘out’ of the museum. The museum would then be defined visually on the model so the movement in and out of the building doesn’t look like pedestrians disappearing and reappearing at a point inside the model. Data on sources and sinks can either be collected by hand, as it has been done previously at bridges, or extracted from readily available information related to attendance at museums. Another method is counting pedestrians from a security camera video feed of the front door. 26
Comment [C13]: Is this defined already?
The concept of ‘disappearing’ and ‘reappearing’ occurs when modeling pedestrian traffic in Venice. Walking is not the sole form of transportation in the city, and many people use multiple forms of transportation throughout a day. If there is no integration between pedestrian traffic and boat traffic in the model, then when a pedestrian ‘gets on’ a traghetto or a water taxi in the model it will look as if someone disappeared from their original position and reappeared somewhere else. To combat this, data can be collected that reflects the number of pedestrians that are getting on and off at each boat stop. Nodes can then be used at each stop in the model to define movement on or off boats. A truly comprehensive Venice traffic model would completely integrate the boat and pedestrian traffic models into one because the various forms of transportation are not independent of one another.
27
Methodology The mission of this project was to collect pedestrian traffic data for the end goal of developing an agent-based modeling system that collects and archives data to effectively predict the behavior of pedestrian mobility streams in Venice. Project Objectives: 1. To quantify pedestrian traffic at key locations 2. To analyze the feasibility of using video based pedestrian traffic counting techniques 3. To organize the pedestrian traffic data into a format capable of helping develop a pedestrian agent-based model 4. To publicize pedestrian traffic data in a visually intuitive format on an online source This project focused on pedestrian movement throughout the district of San Marco in Venice, Italy. A methodology was developed for accurately counting pedestrians, and the feasibility of using video feeds to count pedestrians was investigated. The data that was collected was integrated into an agentbased model developed by the team’s collaborators. Using real time pedestrian counts ensures that the walkers in the model have appropriate timing and destinations. Employing the methodologies that have been developed during this project in future years for the other five districts of Venice will ensure a greater understanding of pedestrian movement in the city. The project occurred from August to December of 2011, with preparatory work during the first 8week term and on site work throughout the latter 8 weeks. The project was limited to gathering data concerning pedestrian congestion, taking into account only the predetermined agent typology. To accomplish this, pedestrians were quantified based on direction of movement and whether the pedestrian was a Venetian or tourist.
28
Figure 7: Area of Study Map
3.1 QUANTIFYING PEDESTRIAN AGENTS To accomplish the project objectives, Team Mobility counted pedestrians at key locations in the area of study. This data was then collected and integrated into a computer model for traffic analysis. To do this, a specific counting method was developed to conduct manual counts based on direction of flow and pedestrian type at key connection points around San Marco. This counting method is to be used by future teams in order to ensure consistent data sets. 3.1.1 Focus Area and Key Counting Locations
The 2010 Venice Mobility team previously analyzed congestion in the San Marco district at ten bridge locations, as seen in Figure ##50. However, the 2011 Mobility team focused on different counting locations, also known as nodes, for the purpose of creating a distinct location for the starting point of the computer model within the San Marco district. The Accademia bridge was the only bridge in common between the two collection years.
50
Amilicar, Marcus, Amy Bourgeois, Savonne Setalsingh, and Matthew Tassinari. Mobility in the Floating City: A Study of Pedestrian Transportation. Worcester: Worcester Polytechnic Institute, 2010.
29
Figure 8: Map of the Ten Counting Locations Used by the B’10 Team
After evaluating a map of the area, the counting locations were determined to take place at the six bridges that connect the two sections of land divided by the Rio San Luca, Rio del Barcaroli, and Rio San Moisè. It was concluded that, because Ponte dell’Accademia is the only bridge on the Grand Canal that leads into the western part of the San Marco district, it should also be analyzed by the team. Counts were also performed at the four traghetto stops in the district along the Grand Canal. These eleven counting locations covered all locations for pedestrians on foot into and out of the western half of the San Marco district. The complete list of bridges and traghetto stops are referenced in Table ##, and the map of each of these is seen in Figure ##. Table 1: Bridges and Traghetto Stops in the Study Area
Study Area Bridges Ponte del Teatro Ponte de San Paternian Ponte de la Cortesia Ponte dei Barcaroli o del Cuoridoro Ponte de Piscina Ponte San Moisè Ponte dell’Accademia
Study Area Traghetto Stops Riva del Carbòn – Fondamente del Vin Sant’ Angelo – San Tomà San Samuele – Ca’Rezzónico Campo del Traghetto – Calle Lanza
30
Figure 9: Google Map of Traghetto Locations and Bridges Locations. Blue Anchors Symbolize Traghetti Stops and Red and Yellow Marker Pairs Symbolize Bridge Locations 3.1.2 Distinguishing Between Agent Types
A useful feature of the pedestrian model is the distinction between pedestrian agent types, such as Venetians and tourists, because each type of pedestrian behaves differently. Venetians have a structured schedule that occurs daily. During the workweek, Venetian pedestrians leave their residence to go to the market, work, or school. The route traveled by locals is usually predetermined to account for the shortest path and time. Tourists are often random in their routes, and travel in a “wandering� pattern. Major tourist sites are often destinations, but they may stop at a shop or restaurant on the way. As a result, tourist movement is less structured. In order to reflect this different behavior in the agent-based computer model, it was important to collect data based on the type of pedestrian. The individuals that were on-site conducting the counts distinguished pedestrians mainly based on visual cues. As previously mentioned, Venetians had more of a direct route, so their pace was steadier, while tourists had more of a random behavior. They often walked with pets or pulled dollies; and businessmen and women or employees were dressed in business attire. Tourists were singled out by whether or not they were holding cameras, or if they were in tourist groups led by a
31
guide. They were more likely to wear leisurely clothing. A complete list of the classifications used is in Table ##. Tourists “Wandering” walking pattern Carries a camera or takes pictures Led by a tour guide Speaks in another language Window shops Looking at a map
Venetians More direct walking pattern Business or uniform attire Briefcase or cart Walking a pet
3.1.3 Counting Method
In order to accurately quantify the flux of pedestrians at bottleneck locations the team utilized a specific counting method, which allowed a quick and efficient method of counting a large number of pedestrians. Once the peak times were discovered (when pedestrian mobility is at its heaviest), manual counts were conducted in the field based on direction of flow. Individuals were stationed at each bridge in clear view of pedestrian flow, with mechanical counters in each hand. Each clicker represented a direction of flow. For example, the clicker in the individual’s left hand represented pedestrians moving away from the counter, and the clicker in the individual’s right hand represented pedestrians moving towards the counter. For fifteen-minute intervals, the individual would click for each pedestrian that crossed the bridge and in which direction he or she moved. Each individual determined a node on the bridge, and clicked for each person to cross that node. For consistency, children being carried by their parent or in carriages, and dogs and other pets were not counted.
32
Figure 10: Example of Counting Based on Direction on Bridge 6
At the end of each fifteen minute interval, the number read on the clicker was recorded into a field form (see Appendix ##) which was later placed into spreadsheets to be submitted for integration into the agent-based computer model. If flow at the peak time was determined to be too heavy for one individual to count, then two individuals were stationed at that location and each individual counted only one direction of flow. This ensured the accuracy of the data collected. To determine the volume of tourists utilizing a specific bridge on any given day, three project members counted tourists while one project member counted total flow for 15-minute intervals for two-hour blocks during the peak volume time. The tourist counts were averaged to account for outliers (if one team member identified a significantly larger or smaller number of tourists) and recorded in database forms. A percentage of tourist attendance at each bridge was calculated by dividing the average by the total number of pedestrians. These percentages were applied to the rest of the bridge data collected by the 2011 Mobility Team and can be seen in Table/Figure ##. The 2010 team performed preliminary field counting to determine the limit of one counter, and found that one counter was capable of recording one direction of flow while distinguishing between Venetian and tourist without being overwhelmed. Their team decided that two counters per 33
location, one per direction, were necessary to reduce the risk of data loss. If a certain time or location was anticipated to have unusually high traffic volumes, the decision was made as to whether or not more than two counters would be stationed at that location. Additionally, to verify the efficiency of the 2011 model and the accuracy of the on location counts, this year’s Mobility team employed same form for our video recording counts which are discussed further in Section 3.2. The counts made by each individual were then collaborated at the end of the time bracket and collected in Excel spreadsheets that were submitted to the collaborators at Santa Fe Complex and integrated into the pedestrian computer model. This data was also converted into a format visible on GIS Cloud for still-time visualizations. Refer to the following section ### for the details on the data collection forms. 3.1.4 Quantifying Traghetti Passengers
The same method for direction-based counting was used for counting at traghetti stops. A clicker in each hand represented the direction of traffic traveling into or out of the study area. The time of when the boat arrived and departed each stop was recorded along with the number of passengers that got on or off the boat. The field form for traghetti counts can be viewed in Section 3.1.5 in Table 2. 3.1.5 Field Forms
To collect all of the data in an organized manner for the utilization of the collaborators, a field spreadsheet template was created. This was used to collect the number of persons that cross through a specific node by type of agent, and in which direction of travel. Refer to Appendix ## for an example of a field form. The same spreadsheet template was used to collect counts through video clips that are discussed in Section 3.2. Table ## shows the columns that were filled out for collection of all on-field data. Table 2: On Site Manual Pedestrian Count Template
Date: Time
Location: Traveling To
Traveling From
Recorder: Count
Number of Tourists
Number of Venetians
The field form for traghetto counts were used to collect the number of passengers that entered or exited the San Marco district using a traghetto. Because traghetti usage is not continuous, arrival and 34
departure times must be recorded. Also, because every passenger on a boat must travel at the speed of the boat, agent mobility patterns are negligible and agent types were not recorded. Table ## shows the template used to record on-field traghetti passenger data. Table 3 On Site Manual Traghetti Passenger Count Template
Date: Arrival Time:
Traghetto: Count Entering San Marco
Recorder: Departure Time:
Count Leaving San Marco
3.1.6 Schedule for Performing Field Counts
For the purpose of having consistent data for a comprehensive computer model of pedestrian flow, the team counted at specific times of day. After determining the peak volume times of and which bridges contained the majority of traffic (as seen in section ###), it was decided that these times would be the best to conduct counts for the model. While data from all times of day would be most ideal, due to the time limitation of seven weeks, the team sought the most crucial data for the framework of the model. Refer to section ## for recommendations on other schedule choices. The team decided that the best time to conduct counts was late afternoon into the early evening, when most people were retiring home from work or most tourists were ending their days or going to dinner. Therefore, a weekly schedule for counting was devised, as seen in Table ##. Table 4: Schedule for Bridge Counts
Bridge Ponte de la Cortesia Ponte San Moise Ponte dell’Academia
Weekday 15:30 – 18:30 15:30 – 18:30 15:30 – 18:30
Weekend 15:30 – 18:30 15:30 – 18:30 15:30 – 18:30
Traghetto stops ran on strict operation schedules that had to be worked around, so time brackets for these counts were developed in order to cover all hours of operation for each stop. Table ## shows the operational hours for each traghetti stop. Table 5: Schedule for Traghetto Stops
Traghetti Stop Riva del Carbon – Fondamente del Vin Sant’Angelo – San Toma San Samuele – Ca’ Rezzonico
Monday – Saturday 8:00 – 13:00 7:30 – 20:00 8:30 – 13:30 35
Sunday 8:00 – 13:00 8:30 – 19:30 Closed
Campo del Traghetto – Calle Lanza
9:00 – 18:00
9:00 – 18:00
The Campo del Traghetto to Calle Lanza traghetto was closed for work while the team was taking counts, therefore, no was collected for that stop.
3.2 DETERMINING VIDEO SURVEILLANCE FEASIBILITY Looking to provide as accurate as possible a computer traffic model for the city of Venice, a ‘proofof-concept’ was developed in order to test the feasibility of using a video surveillance system to collect pedestrian traffic data. The goal of the ‘proof-of-concept’ for the Mobility Team was to provide a variety of video-feed samples that represent the complexity and variety of pedestrian traffic in Venice. These video feeds provided an appropriate and comprehensive data set to test the feasibility of using remote counting techniques coupled with video surveillance feeds as an alternative to the manual field counts that have been the basis for WPI Mobility Teams’ counting over the past several years. 3.2.1 Filming Scenarios
The video feeds collected were each fifteen minutes in length to provide continuity among pedestrian traffic data collected. The feeds covered a variety of scenarios often seen in Venice so that the proof-of-concept could reflect the range of possible scenarios. The variety of camera feeds also demonstrated several different camera angles and orientations. The various orientations served to provide a means to determine which camera orientations provided a viable frame of reference for software and video based manual counts. Below are the scenarios that feeds were collected for: Camera Angles Bird’s Eye View Bird’s Eye View Bird’s Eye View Bird’s Eye View Horizontal Straight On (Directly facing the traffic flow) Horizontal Straight On (Directly facing the traffic flow) Horizontal Perpendicular (Facing perpendicularly to the traffic flow) Horizontal Perpendicular (Facing perpendicularly to the traffic flow)
36
Scenario High Volume of Traffic (Day Time) Low Volume of Traffic (Day Time) High Volume of Traffic (Night Time) Low Volume of Traffic (Night Time) High Volume of Traffic Low Volume of Traffic High Volume of Traffic Low Volume of Traffic
Comment [FB14]: Is it possible to grab a few of these video frames and insert them into the paper…it would be very helpful…
3.2.2 Camera Set Up
To collect video feeds at bridges, a GoProTM HD Hero Camera was used. For the ‘Horizontal Straight On’ and Horizontal Perpendicular’ camera angles the camera was set up using a simple tripod. The tripod was then placed in a spot where it could collect the video feed and not impede traffic. For the ‘Bird’s Eye View’ camera angle, the camera was attached to a 12 foot boom measured from the base of the rigging apparatus (rig). The rig was then securely fixed to the side of the bridge being counted using an industrial grade strap. Additional lashings were tied using 1/8 inch rigging line to provide extra stability and ensure structural integrity throughout the recording process. The rig was attached in such a spot so that it would not impede traffic but would still provide a bird’s eye view of the spot on the bridge where there was the most constant flow and width was minimal. These spots directly correspond to assigned nodes on the Study Area map. The camera lens was aimed parallel to traffic flow. All of the camera angles used the ‘r4’ video resolution mode on the HD Hero camera. This setting provided the most vertical viewing area with the maximum overall view. The video was collected in HD 960p resolution. 3.2.3 Statistical Comparison of Manual Couting Methods
Once collected the video feeds could be counted remotely and then used to verify the field counts that were conducted simultaneously with the collection of the video feeds. Remote counts were conducted using the video by analyzing the feed frame by frame. This process was conducted to give what could qualitatively be considered the most accurate count. This assumption was made based on the fact that in remote counting time is no longer a factor as is the case with field counts. In field counting the counter only gets one chance to collect an accurate data set, but in remote counting the counter gets as many tries as needed and can even conduct multiple counts to conduct a statistical analysis if necessary. The remote video counts were then compared to the field counts to determine how precise the two counting methods are.
3.3 ANALYZING AND VISUALIZING COLLECTED DATA To provide a streamlined method of inputting data from the field into the final model, all the data was reorganized into Database forms. These forms contained a format which was cohesive with the programming of the model. In addition, these forms would allow for the bridge counts to act as a 37
Comment [FB15]: Do you have a picture of the set-up…it would be helpful..
verification method that the final model is indeed accurate once completed. The reformatting consisted of organizing and explaining data in terms of nodes, geographical points which have the ability to store parameters within them. These nodes exist for both locations which serve as sources of pedestrian flow, as well as attractors, or pedestrian destinations. In addition, information was acquired from various alternate sources including field counts from past years, as well as the Census Statistics Office. 3.3.1 Nodular Formatting
To ensure that the agent-based model was performing as anticipated, the team came up with a usable format for tabulating the collected data for the programming requirements of the collaborators. Nodes, the location based entities or geospatial points in our model environment necessary for directing traffic flow accordingly, were created on the study area map, based on nodes already in existence from past studies. These nodes would aid in the directional flow of pedestrian traffic within the model, creating constrictions on how many pedestrians travel from one location to another. Within the model, nodes exist exclusively along pathways, and as a series of points define ‘edges’ which the agents use as the pathways themselves. Nodes within the model each are titled by a number of approximately five digits in length, and contain many parameters which determine how many pedestrians cross daily, have already crossed, and will cross in the future. In order to simplify this nodular premise during field counts, key nodes of study were given simpler letter based names. For example, a pedestrian crossing Bridge 6, could cross from node L on one side of the bridge, to node K, and head towards the Ponte dell’Accademia, for visa versa for the direction of the Piazza San Marco. For locations such as residential areas, places of work, schools, and areas which to tend to attract short-term visitors, nodes were also created because these are pertinent to the creation of the most accurate model possible. Much of the information regarding these sources and attractors was based off of data received from the Census Statistics Office. The parameters within the nodes define the majority of the functions of the model. For example, nodes have certain elements within them that define what types of people venture to them, based on algorithms that define ‘nodular attraction’. 3.3.2 Rules of Attraction
For the purpose of programming, the most useful format for the data within these spreadsheets was to leave the data in its rawest form, as the counts themselves, in addition to determining the ratio of local and visiting population which frequent these nodes daily. The model originally contained 38
random walkers which were then constricted by different rules for each agent type. These rules contain nodular attraction which, based on a probability, would draw or repel pedestrians. In addition, rules were added to create the chronologic effect of a typical day within the city. This included having pedestrians wake up at various times in the morning at their source node, travel to their respective destinations throughout the day, and end at the same source at various times during the evening. For example, the average 40 year-old Venetian would awake early in the morning and take a direct route to his/her place of labor, spend time there until travelling home, where they may run errands and stop at markets, or other stores on their way back to their residence. Tourists would likely behave much differently, starting their day later, either at an entrance to the city of Venice or a lodging facility, and travel for much of the day, wandering between various sites, and finally returning to their point of origin. Each location node would have a different attractive force on each of the two agent types. This force of attraction, which was defined as FA, required multiple parameters to be considered in order to accurately model reality. These parameters included the individual’s desire to venture to a destination, as well as the individual’s distance from that destination. Similarly, the electromagnetic attraction and repulsion between subatomic particles is defined by two parameters, including distance and charge. By relating the criteria of desire to electromagnetic charge, a formula which defines FA in a similar manner as Coulomb’s Law was used.
Figure 11 - Coulomb's Law
In order to ensure an accurate number of each agent type arrived at every destination in the model, the “charge” of the location node was determined based on a ratio how many people had already arrived versus the number of people which frequent that destination. This ratio constantly changed throughout the day and existed simultaneously for each location. For modeling purposes, each node then required data to define the number of daily venetians, daily tourists, and continuously calculate the number of Venetians arrived, and tourists arrived. Furthermore, the two ratios would also exist as follows:
39
1
Figure 12
1
Figure 13
These two equations explain how, for the purpose of the model, the relationship between the attractive strength of a node had a negative correlation to the arrival of pedestrians. Furthermore, the attractive force, FA, was also modified by distance. To accommodate this, the r2 value was determined by the relative distance between the pedestrian and destination, based on the route of travel. This was important to implement, because a tourist that wishes to go sight-seeing, is more likely to go first to destinations that are both desirable as well as in the vicinity. After combining all these elements, the final equations which describe the relationships between each node and agent are as follows:
Figure 14 - Force of Attraction
In order to implement all the data collected during field counts, which led to the development of the daily pedestrian statistics, Excel spreadsheets were submitted to our collaborators. These spreadsheets, by utilizing these nodular locations and the relationships described, were integrated into the pedestrian model in a format compatible with the programming language HTML5, which was used to create the model. 3.3.3 Census Tracts and Statistical Data
The remaining nodes within the area of study required additional data not provided through pedestrian counts. These nodes included many destinations of the model, such as places of work, as well as sources, such as residencies and various types of lodging facilities. In order to fulfill the requirements of the model and create probability data to appease the rules of agent attraction, the parameter which describes the total number of daily attendees needed to be discovered.
40
Fortunately, census tracks are publicly available by request, and have the added precision of breaking the city down, not only by its districts, but also into almost four-thousand sections. This sectional breakdown allowed for a much more precise organization of data. These tracts contain information regarding the population, with gender and age breakdown, as well as numbers of both residencies and businesses which exist in each section. This supplementary data was organized into a spreadsheet form in order to apply it to the pedestrian model, where it would satisfy the remaining parameters for determining the attractive strengths of many locations, as well as the number of Venetian agents which would start and end their day in each particular section. In order to create a more accurate model, the ages of the local citizens was taken into account, and based on observation, would behave differently in regards to travel. For example, a resident between the ages of 15 and 19 was likely to attend school in the morning, whereas a Venetian of twice that age would be travelling to a place of occupation.
3.4 PUBLICIZING DATA AND RESULTS To ensure this project can be furthered by future Mobility teams, the 2011 team publicized data utilizing several different sources. Using visual aids, called deliverables, the team was able to present pedestrian traffic data through informative means, including videos, graphs, and maps. The methodology described throughout this section was implemented with the purpose of being used by subsequent project teams to expand upon the pedestrian model that demonstrates walkers’ movement throughout the entire city of Venice. To aid in expanding upon the model, all data and information can be extracted from the team website at sites.google.com/site/ve11mobi. There are separate sections for data, video clips, and graphics that display all of the visual results from the 2011 project, and the information that is in each section on the website can be seen in Table ##. Information
Location on Website
41
3.4.1 Deliverables
Deliverables are informative visual aids that aptly demonstrate the data collected and analyzed throughout the project. Mobility’s deliverables included graphs, video feeds, a time-lapse video, and a study area map. The graphs can be seen in sections ##, and are further analyzed in section ##. The video feeds that were taken at select bridges during the project contributed to the final pedestrian model, as described in section ##. A time-lapse video of the pedestrian flow over the ##
Comment [CS16]:
Bridge can be seen in the “Deliverables” section of the 2011 Mobility website. This video shows the difference in the magnitude of people traveling over the bridge between the hours of ## and ##. The study area map, as seen in Figure ## is a Google Map of the San Marco district, with different “layers,” or data sets, that consist of sources and sinks, including residential districts, hotels, traghetto stops, hotels, museums, churches, schools, places of employment, and other key locations. Figure ## illustrates many of these types of sources and sinks.
Figure 15: Sources and Sinks
Ideally, this visual aid will allow the public to see the congestion locations and reconsider their route across Venice, taking into account the most congested areas as seen on the deliverable map. 3.4.2 Furthering a Pedestrian Model
The end goal of this IQP was to collect data in such a way as to further the development of a pedestrian traffic model that simulates the movement of people throughout the city of Venice. To do this, the team developed and followed the methodology outlined in the previous sections, and compiled all of the collected data. This enabled the collaborators to use accurate data for the 42
Comment [C17]: Graph or some compilation of data.
foundations of the model. The Results and Recommendations sections describe in detail how the project can be expanded upon by future project teams.
43
Results and Analysis 4.1 DETERMINING CONSTANTS Once the methodology was implemented, the team analyzed and compared each collected data set to view trends that appeared throughout the data range. These trends appeared consistent throughout the data, and can be considered constants for the scope of Venetian pedestrian studies. The constants are summarized in Table ## and explained in depth in the following sections. 1 2 3 4 5 6
There are peak times Peak times are consistent day to day Weekday peaks are of similar magnitude Weekend peaks are of similar magnitude Specific bridges carry the majority of traffic flow Secondary bridges carry an insignificant flow
4.1.1 Peak Times and Bridge Priority
Pedestrian traffic flow over bridges is not constant throughout the day. There is an ebb and flow as time passes, with a greater number of people crossing a bridge at certain times, and a lesser number at other times. The high pedestrian flow times are of greater significance, because these are the times when traffic is at higher risk of congestion. For the purpose of this project, the time frames containing high flow rates were considered the “peak times� and were the times given the most focus. Below is a graph of the weekday peak times for Ponte del Teatro. Data was collected on three separate days and, by evaluating the graph, the peak time was determined to be in the early afternoon, more specifically from 13:00 to 14:00.
44
Weekday Peak Times Ponte del Teatro 70
60
50 31‐Oct 1‐Nov
40
2‐Nov 31‐Oct
30
1‐Nov 2‐Nov
20
10
0 11:45 12:00 12:15 12:30 12:45 13:00 13:15 13:30 13:45 14:00 14:15
The same appears to hold true for Ponte de San Paternian as seen in Figure ##.
Weekday Peak Times Ponte de San Paternian 60 50 31‐Oct 40
1‐Nov 2‐Nov
30
16‐Nov 31‐Oct
20
1‐Nov 2‐Nov
10 0 11:15 11:30 11:45 12:00 12:15 12:30 12:45 13:00 13:15 13:30 13:45 14:00
45
These two bridges, however, do not experience high volume flow throughout the day. The maximum number of pedestrians that cross the bridge in a fifteen-minute interval does not exceed 60 for either bridge. Figure ## below shows a high traffic bridge, with a maximum of about 500 pedestrians crossing in a fifteen minute interval. Ponte del Teatro only receives a little more than ten percent of the maximum number of pedestrians Ponte de la Cortesia receives. Therefore, bridges like Ponte del Teatro, Ponte de San Paternian, Ponte dei Barcaroli o del Cuoridoro, and Ponte de Piscina were rendered secondary bridges. And Ponte de la Cortesia, San Moise, and Ponte dell’Accademia were the primary bridges. For the duration of the project, the primary bridges received the most focus for data collection. This breakdown is summarized in Table ##. Primary Bridges Ponte de la Cortesia Ponte San Moisè Ponte dell’Accademia
Secondary Bridges Ponte del Teatro Ponte de San Paternian Ponte dei Barcaroli o del Cuoridoro Ponte de Piscina
Primary and secondary bridges exist because of the destinations each bridge leads to. A bridge that was between a tourist destination and a popular view of the city or led to a street with shops and restaurants had many pedestrians crossing. Alternatively, a bridge that led to a small alleyway with Venetian apartments was much less traversed. Below is a graph of the peak time for Ponte de la Cortesia, and it displays that the peak times are a little later in the afternoon, from 13:00 to 14:00.
46
Weekday Peak Times Ponte de la Cortesia 600 500 400
1‐Nov 2‐Nov
300
14‐Nov 1‐Nov
200
14‐Nov 100 0 11:15 11:30 11:45 12:00 12:15 12:30 12:45 13:00 13:15 13:30 13:45 14:00
The peak time for Ponte San Moisè occurs slightly earlier, from 12:30 to 13:15, but it experiences the highest traffic levels, with a maximum amount of pedestrians in a fifteen-minute interval being about 900.
Weekday Peak Times San Moise 1000 900 800 700 31‐Oct
600
2‐Nov 500
14‐Nov
400
18‐Nov
300
14‐Nov
200 100 0 11:30 11:45 12:00 12:15 12:30 12:45 13:00 13:15 13:30 13:45 14:00
47
Below is a graph for the peak times at Ponte dell’Accademia. After evaluating data collected by the 2010 Mobility team, it appeared that Ponte dell’Accademia was most crowded during the evening hours. Therefore, counts were conducted later in the day. According to the graph, Ponte dell’Accademia achieves its maximum level of traffic between 16:00 and 17:00, much later in the day than the other bridges evaluated.
Weekday Peak Times Ponte dell'Accademia 500 450 400 350 300 15‐Nov 250
22‐Nov
200
15‐Nov
150 100 50 0 15:30 15:45 16:00 16:15 16:30 16:45 17:00 17:15 17:30 17:45 18:00
Once counts were conducted throughout the day and the resulting data was analyzed, peak times were discovered. The average time frame of the peak times for three bridges can be seen in Table ##. Conducting manual counts for a segment of time before and after the peak times allowed a clear bell curve to be seen, with the quantity of pedestrians crossing the bridge gradually increases until the apex, then gradually decreasing. Initially counting throughout the day also provided the team with the low flow times, which could then be eliminated from the counting time brackets. The low flow timeframes could be deleted because they did not offer any statistical value in terms of the pedestrian model.
48
Bridge Ponte de la Cortesia Ponte San Moisè Ponte dell’Accademia
Evening Peak Time Bracket 12:00-14:00 12:00-14:00 16:00-17:00
The early morning times did experience smaller peaks caused by locals traveling to their places of employment and schools; tourists were rarely walking the streets at this time. These peaks were not as significant as the afternoon/evening peaks, which were contributed to by both Venetians and tourists. Venetians were traveling home from work and school, and tourists were exploring the city and sightseeing. The difference in evening peak times between bridges has a logical explanation, especially in terms of the Ponte San Moisè and the Ponte dell’Accademia. As seen in Figure ##, the Ponte San Moisè (location B) is between a major tourist destination, Piazza San Marco (location A), and Ponte dell’Accademia (location C), which crosses the Grand Canal and is a main method of getting from one district to another. If tourists decide to leave St. Mark’s Square and travel to Ponte dell’Accademia, the easiest route takes them over Ponte San Moisè and on towards the Accademia. As the pedestrians traverse this pathway, the peak time occurs at each bridge in turn, with the Accademia peak being delayed by the time it takes pedestrians to walk from San Moisè to the Accademia. Therefore, the Ponte dell’Accademia peak time is later than that of the Ponte San Moisè. Similarly, if locals work in the district of San Marco and live in the district of Dorsoduro, across the Grand Canal, one option for the commute home is the Accademia bridge.
49
Determination of peak times and primary bridges allows counts to be focused on the three primary bridges and at specific time brackets without missing critical data. 4.1.2 Weekday Peaks are of Similar Magnitude
As seen in the graphs above (see Figures ##) of weekday peaks for each bridge, very frequently peaks were of similar magnitude varying from weekday to weekday. After analyzing the data compiled into spreadsheets, it was determined that weekday peaks have counts that were in fact within 5% alpha significance level. This is because Venetians will follow the same path to and from work or school every weekday, and tourists will follow similar paths to and from their tourist destinations, aside from the small percentage that get lost. Therefore, weekdays can all be treated the same and counts can be conducted for a general weekday. 4.1.3 Weekend Peaks are of Similar Magnitude
Due to time constraints, we were unable to obtain enough information to either confirm or deny the assumption that weekend peaks are of the same magnitude. The data we were able to collect for comparison was within ten percent significance of each other, but further data collection and comparison would have to be conducted to confirm the validity of this assumption. Therefore, this assumption was rejected for this project.
50
4.1.4 Overall Assumptions Results
Of the six assumptions established at the start of this term, all of them were accepted as true for this project except for the assumption that weekend peaks are of similar magnitude. Therefore, the constants that were established for the purpose of collecting data for this project were that there are in fact peak times for each bridge, that peak time are consistent from day to day, that weekday peaks are of similar magnitude, that specific bridges carry the majority of traffic flow, and that secondary bridges carry insignificant traffic flow.
4.2 COLLECTED DATA 4.2.1 Counting Methodology
The methodology 2011 Mobility used for conducting field counts was outlined in Section 3, and it was successful in quantifying the number and type of pedestrians traveling throughout the study area. As outlined in Section ##, when the team differentiated between Venetians and tourists while counting, three team members counted the quantity of tourists crossing the bridge in each direction, and the fourth member counted the total number of pedestrians. The three resulting tourist counts, as seen in Appendix ##, were comparable and the average was used in data analysis. The results from the tourist versus Venetian counts can be seen in the pie charts below. Since field counts were taken at each bridge for the morning and evening Venetian commute times, the two timeframes were compared side by side. The agent breakdown for Ponte de la Cortesia for an average weekday is demonstrated in Figure ##. This clearly shows the difference between the number of tourists and Venetians crossing the bridge during the morning and evening hours. Though the quantity of tourists increased over the course of a day, Venetians outnumbered tourists on this bridge during the commute hours.
51
Agent Breakdown Average during Morning Peak Morning
Agent Breakdown during Evening Peak
Tourists 10%
Average Evening Tourists 26%
Average Evening Venetian s 74%
Average Morning Venetians 90%
Figure 16: Agent Beakdown for Ponte de la Cortesia
The results from tourist versus Venetian counts for Ponte San Moisè can be seen in Figure ##. As was the trend at all bridges, the number of tourists traversing bridges increased as the day progressed. Unlike Ponte de la Cortesia, however, the number of tourists outnumbered Venetians in the evening peak.
Agent Breakdown during Morning Peak
Agent Breakdown during Evening Peak
Average Morning Tourists 30%
Average Evening Venetia ns 35%
Average Morning Venetians 70%
Average Evening Tourists 65%
Figure 17: Agent Breakdown for Ponte San Moisè
Figure ## demonstrates the agent breakdown for Ponte dell’Accademia, for which the trend was similar to Ponte de la Cortesia. Venetians outnumbered tourists during the morning and evening commute times, and the number of tourists increased throughout the day.
52
Agent Breakdown during Morning Peak
Agent Breakdown during Evening Peak
Average Morning Tourists 8%
Average Evening Tourists 28%
Average Morning Venetians 92%
Average Eveing Venetian s 72% Figure 18: Agent Breakdown for Ponte dell'Accademia
The number of tourists increased as the day progressed because more tourists began to travel to their destinations later in the day, and as the peak time arrived, the highest number of tourists was traveling throughout the city. 4.2.2 Traghetti Usage
While the intention was to count passengers at four traghetti stops in the San Marco district, the Campo del Traghetto-to-Calle Lanza crossing was out of operation during our designated time. Therefore, counts were performed at the other three traghetti crossings in the district and results are given for each traghetto’s full hours of operation collected over several days. The following map displays the comparison of passenger usage at each traghetto and the proportion of passengers entering versus the proportion of passengers leaving the San Marco district.
53
The orange bars on the graph represent the total number of passengers that used each traghetto stop in that day, illustrating that Sant’Angelo was more frequently used than the others. However, it also operates for more hours throughout the day. The charts displaying the traghetti usage over time for each stop can be view in Appendix ##. The graphs illustrate that the crossings are used mainly in the morning to enter San Marco. The greater level of usage in the morning to enter San Marco district over exiting the district is a notable aspect. This is most likely because the district is heavily populated with residential areas along the Grand Canal, as seen in Figure ##. Locals are using the traghetti to get back home after working outside of the district, or after they drop off and pick up their children from school.
54
RESIDENTIAL LOCATIONS 4462 RESIDENTS
Figure 19 Gradient Map of Residential Locations
The following map shows the work locations in the San Marco region. There are less work locations in the district, therefore most of the residents must commute out of the area for work.
55
WORK LOCATIONS Figure 20 Gradient Map of Work Locations
Overall, the usage of traghetti usage was less than expected, considering most locals would more than likely not want to use the major bridges that tourists frequently cross. An observation made during counts was that almost every passenger was a local, indicating that most tourists have not discovered these convenient and inexpensive gondola boats. 4.## Bridges As discussed in the Analysis Section 4.###, Team Mobility conducted manual counts at the three primary bridges after the initial counts were taken. Even between these three bridges, however, there was a significant difference in the quantity of pedestrians crossing. As seen in Figure ##, Ponte San Moisè has the highest flow on average weekday of the three bridges, while Ponte de la Cortesia and Ponte dell’Accademia have a significantly lower flow.
56
The figure above also clearly demonstrates the peak times, which were discussed in Analysis Section 4.###. As time progresses throughout the day, more and more pedestrians cross the bridges in their travels to work, school, or tourist destinations. The amount of walkers crossing the bridge eventually reaches an apex in the afternoon or evening, and decreases for the rest of the night. This is repeated daily, and there is a similar trend on weekends. 4.2.3 Complete Study Area
Once the data was compiled for the Actv stops and collected for the tragetti stops and bridges within the study area, the datasets were analyzed. Figure ## demonstrates the total flow for an average day at each of the entrances and exits for which the team gathered data within the study area.
57
Figure 21: Total Number of Pedestrians Entering and Exiting the Study Area on an Average Weekday
When the number of passengers utilizing the Actv and the traghetti were compared, it was evident that the Actv carried almost four times as many passengers. The number of pedestrians using bridges clearly outnumbered the boat transportation because of the higher quantity of bridges compared to the boat stops, the ease of access, and the lack of a utilization fee. There was also a consistency in the number of people entering and exiting the study area throughout the Actv, bridge, and traghetti data. For each type of data, there was a greater number of people entering western San Marco than leaving. As discussed in Section ##, this was because of the Venetians traveling to work or school in the district, and the high number of tourists traveling between the Piazza San Marco and Ponte dell’Accademia. 4.2.4 Determining Video Surveillance Feasibility
There is no statistical difference between each team memberâ&#x20AC;&#x2122;s counts in both high flow and flow scenarios. There is also no statistical difference (alpha value <.05) between the video manual counts and the manual field counts in any of the tested flow or camera placement scenarios. This presents the conclusion that counting remotely through video feeds is feasible and will provide accurate pedestrian traffic flow data. See appendix ## for comparison analysis.
58
4.3 COMPILED DATA After performing field counts for pedestrians at bridges and for passengers for the traghetti, this data was compiled with the data received from various other sources. These sources include census tracts from the Statistics Office of Venice, and ridership data from the ACTV. These exterior data sources were implemented in the model; in addition, some basic examination of trends in the data was also done. 4.20.1 Census Tracts
Data regarding the population of the city of Venice was determined to be vital in order to ensure the computer model produced was as accurate as possible. This data would answer the fundamental questions underlying the agents modeled, including “who are we modeling” in addition to “how many?” The data from the statistics office was organized into spreadsheets and contained information for specific regions of the city of Venice. Within the spreadsheets was valuable information regarding details about the population of the city, including the number of males, females, employed individuals, number of firms or businesses, as well as an age breakdown by intervals of five years. This data was extremely important in creating properties for our virtual environment, as well as mapping out origins and destinations for the local population. Below is a GIS layer, which indicates the number of residencies within each given tract, or section, of the district of San Marco.
59
Figure 22 - San Marco Residencies
In this diagram, the different gradients signify different numbers of residencies within them. Specifically darker variants of green have a great number, with the darkest having between 26 and 113 residencies. Areas white in color have no residencies within them. Similar evaluation was conducted in regards to the number of businesses within each section of the district of San Marco.
Figure 23 - Employers within San Marco
Though this census data gives us a better understanding of how the local residents might move around within the district of San Marco, there are also a number of factors left unexplained to this 60
point. To better understand the movement of the local Venetian population, those which commute into and out of this district for the purpose of work must also be accounted for. 4.20.2 ACTV Ridership
ACTV is the local company which provides the area with Mass Transit services in the form of vaporetti. The company transports thousands of passengers around the city of Venice each day, including both tourists and Venetians. There were three ACTV stops within the scope of the area of study for this project, which cover two of the many lines offered by the company, lines 1 and 2. Line 1 stops at approximately 20 stops and focuses upon the City Center by travelling around the Grand Canal, furthermore, Line 2 is typically used as an express line, as it stops at many less locations between San Marco and Piazzale Roma. Along these lines, the three stops focused upon were Sant’ Angelo, San Samuele, and Santa Maria del Giglio. To observe the way that mass transit affects the district of San Marco, ridership data from the year of 2009 was acquired from the Statistics Office of the Commune of Venezia. From this we were able to determine the average usage of the ACTV per day, as well as the typical number of people who both board and exit ships at our focus area each day. The data for each of the stops under study was then placed into
Early Morning Morning Noon Afternoon Evening
7:00 – 9:00 9:00-11:00 11:00-13:00 13:00-17:00 17:00-19:00 Table 6 - ACTV Hour Groups
spreadsheets, and graphed for an
analysis. The data was previously grouped into sections of the day, each consisting of a few hours, as seen in the Table to the right. To convert the data into a common form which could then be analyzed and studied for trends, all the raw passenger counts were converted into an hourly format, passengers per hour, and organized by their activity. The final line graph for the popular Sant’Angelo stop can be seen below.
61
Axis Title
Hourly Passengers: Sant'Angelo 550 500 450 400 350 300 250 200 150 100 50 0
Hourly Entering Hourly Exiting Hourly Passengers
7:00‐9:00
9:00‐11:00
11:00‐13:00 13:00‐17:00 17:00‐19:00
Figure 24 - Passengers per Hour at Sant'Angelo ACTV
One trend that is consistent between all of the stops studied was determined through statistical analysis. We found that a significantly greater number of passengers arrive in the morning than depart at that time, and the opposite becomes true in the evening. Due to a large number of people commuting into the area in the morning, and leaving in the evening, we determined that this would be partially due to tourists in the area, but also due to local Venetians who live outside the district of San Marco entering and exiting during their commuting times to and from work. In addition the number of passengers for each stop was compiled and compared for future analysis with other types of transportation. It is important to note that Sant’Angelo and Santa Maria del Giglio are stops on Line 1, which has been described as popular with both the local population, as well as with tourists, while the San Samuele stop is located on Line 2, an express line that helps tourists travel from points of entrance to large attractions.
62
Full‐Day Comparison of ACTV Stops 5000
4330
Number of Passengers
4500
Full Day Disembarkment
4000
Full Day Boarding
3500
Totals Full Day
3000 2500
2243
2116
2087
2000 1500
1153
1000
1131
963 523 608
500 0 Sant'Angelo
Santa Maria del Giglio
San Samuele
Figure 25 - 2009 ACTV Ridership
All of this information has helped describe the movements of pedestrians into and out of the District of San Marco; moreover, information in regards to the origins and destinations of the many tourists of the area will be required to model their behavior accurately.
4.4 PEDESTRIAN TRAFFIC COMPUTER MODEL The autonomous agent model is a resource for higher understanding of pedestrian mobility in Venice. This model was coded in HTML5 language for reasons of its flexibility as a language, as well as its ability to be easily reached by the public through most any current internet browser. This language allowed for the necessary logic to be developed and implemented into the model via a collaborative effort between our team, and Cody Smith, an Intern from the Santa Fe Complex who is both knowledgeable and experienced with the language. 4.4.1 Environmental Framework
The first step in ensuring the model’s accuracy was to create a virtual environment for the agents to exist and occupy. Our design utilized maps created through geospatial information systems, known as GIS. The GIS mappings constructed our virtual environment of the model used accurate archived data collected throughout the years of studies done within the City of Venice. Most important for
63
this environment was the accuracy of the locations of pathways within the city. Both a satellite image as well as the pedestrian path layer of our GIS map can be seen below.
Figure 26: Comparison Between Satellite and GIS Maps
The conglomeration of these pathways, as stated previously, are made up of edges, which are constructed as a series of nodes, the building block of geospatial information. Each of these nodes has an identification number to be referenced and knows its distance from other nodes within the network. This setup is known as flood-fill networking. The GIS layer containing the pedestrian walkway information contains enough detail to properly map all bridges, thruways, and dead ends within Venice; furthermore, its contents are updated every few years, making it both a versatile and continuously accurate modeling environment. 4.4.2 Agent Movement
The second main component of the model’s function is contingent upon the behaviors of the agents within it. These agents serve the purpose of providing a visual representation of the various types of their pedestrians counterparts which move about the city on a daily basis. Each of these agent types will behave differently within the model and travel to different locations at different times based on their decisions made throughout the day. These trip planning decisions have many factors involved, and will be made autonomously based upon the ‘rules of attraction’ described in Section ##.##.##, as well as decisions made on-the-fly as they cross particular nodes. In addition, to improve functionality, the notion of chronological behavior was also implemented to aid in this decision making process and augments these rules of attraction based upon the current time within the model.
64
Figure 27: Morning Commuting in the ModelÂ
The behaviors of different pedestrian agent types were modeled through their individual probabilities of being in attendance at one of many different types of locations. Certain trends were taken into account, such as the commuting of the local population from their residence to the work in the morning, while tourists are simultaneously either getting breakfast at a food establishment, or not out and occupying the streets just yet. In the figure above, the two types of agents, locals and tourists, are represented by green and red dots on the map, respectively. On the right side of the screen, a time based plot, as well as pie charts containing the current locations of all pedestrians, can be found for more quantitative information regarding the conditions present in the model. As the day progresses, the crowds move to time-appropriate locations. Below, another image taken at approximately noon in the model shows most agents have ventured to food establishments for lunch.
65
Figure 28: Lunchtime for Agents 4.4.3 Final Model Functionality
The autonomous computer model provides many functions that allow for a better understanding of the movement of pedestrians within the city of Venice. First and foremost, the model brings together data from a wide variety of sources and organizes it into one single location, and presents it in a visually appealing and user-friendly manner. By compiling this data in one spot and using algorithms to determine behaviors, the model allows for any gaps in data to be accounted for. Once this has been established, the model serves two main purposes, based upon the time of reference. Within the present, the model serves the purpose of providing an accurate visualization of the dynamic movement of those within the city right now; serving the important purpose of filling in the gaps of data that would be otherwise missing. Secondly, the data involved in the processing of this model can be extrapolated for predictions of how movement in the city will behave in the future. By assessing data, and making these predictions available to the city, proper preventive measures can be taken to alleviate some of trafficâ&#x20AC;&#x2122;s stress upon the cityâ&#x20AC;&#x2122;s infrastructure and the people within it.
66
4.5 PUBLICATION At the conclusion of the project, the 2011 Mobility team publicized all results, datasets, and helpful information in a variety of sources. The publicized information can be used by future groups or organizations interested in Venetian mobility. 4.5.1 2011 Mobility Website
The website that was maintained throughout the duration of the 2011 project contains links to everything that was applicable to the project. A list of the different items and the location on the website can be seen in Appendix ##. The website is available at this link. The website is a centralized database of all information pertaining to mobility in Venice that the 2011 team utilized, collected, and compiled. 4.5.2 Youtube
The videos that were recorded during bridge counts and used for both the methodology verification and proof of concept are available on the 2011 Mobility Youtube account here. There are two types of video samples: qualitative and quantitative. The quantitative videos are time lapses that show the movement of pedestrians over time, as described in Section ##.##. The qualitative video clips are positioned in a viewpoint so as to be able to see pedestrians crossing a bridge. Camera viewpoint analyses can be seen in Section ##.##. The description of each video sample is included in Appendix ##. 4.5.3 General Mobility Website
In addition to up keeping the 2011 Mobility website, the team also developed a general Venice Mobility website, which is available at this link. This website links to all of the past WPI IQP mobility studies, and will be further developed by future teams. It will be a comprehensive source for all mobility-related information, research, and results for the city of Venice. 4.5.4 2011 Mobility CD
As a supplement to the online sources, the 2011 Mobility team published all of the results and helpful information on a CD that can be used by future teams investigating Venice mobility. Included on the CD will be all of the compiled information, collected datasets, and GIS map layers.
67
Recommendations 5.1 PEDESTRIAN MOBILITY EVALUATION RECOMMENDATIONS More pedestrian counts can be performed to improve upon the computer model and gain a more comprehensive understanding of pedestrian mobility in Venice. A continuation of the studies of the San Marco district as well as the other five districts of the city will supplement the data collected this year and in previous years for the ultimate goal of creating a model that will be used to prevent traffic congestion. 5.1.1 Continued Raw Data Collection Additional raw data should be collected and compiled to supplement the counts. Each census report and more ACTV data should be archived in the same public location as the data collected this year. More data concerning hotel occupancy, and museum attendance should also be compiled. Also, various stores have automated counters at their entrances that count the number of people who enter and exit. This would be useful information to obtain to also archive.
Figure 29 A counter outside of a stationary store in San Marco. It has the ability to count the number of people entering and exiting the store.
5.1.2 Expansions of Bridge Data Collection The more data that is collected at bridges, the greater the programâ&#x20AC;&#x2122;s ability to model traffic flow. Therefore, further counting should be performed at the three bridges utilized this year. Additionally, counts should be conducted at the other bridges in the San Marco district and throughout Venice. Data should also be collected at more times during the day. This yearâ&#x20AC;&#x2122;s team chose to collect at peak volume times. Since WPI projects are restricted to the beginning seven weeks of the tourist off68
season, collecting more data and at other times of the day would provide a more accurate extrapolation for traffic during the peak tourist months. To provide the model with more data, other times should be considered. Continued studies into the proportions of Venetians and tourists that utilize specific bridges will help provide insight into the bridge choices each agent type is likely to use when there are more than one option. More detailed agent distinction will provide more accurate behavioral patterns for the model. Distinguishing between day tourists and overnight tourists would be very beneficial, each type has a different origin and would behave differently throughout the day. Overnight tourists tend to visit the major attractions and spend more time at each attraction. Day tourists tend to visit secondary attractions due to the city’s accessibility, and visit more attractions, due to the individual’s time limitations. 5.1.3 Intersection of Traghetti and Pedestrian Traffic Further studies of the traghetti stops analyzed by this project should be conducted. The rest of the traghetti stops along the Grand Canal should additionally be studied to better thread together the connection between traghetti transportation and pedestrians on foot. Other useful information that should be collected is the percentages of locals and tourists who use the traghetti. A thorough analysis into whether or not traghetti are a critical mode of transportation for pedestrians would be of use to the sponsors of this project at the City of Venice Department of Mobility. 5.1.4 Study of Other Situations This year’s project team decided to count during ideal circumstances, without unusual weather conditions such as heavy rain, extreme cold, or thick fog or during aqua alta, high tides. These types of weather conditions impair traffic mobility and should be studied in order to gain a complete representation of Venetian mobility in the model. Other mobility impairments that data should be collected for would be pedestrians with handicaps, strollers, and carts. These features slow down an individual’s pace and consequently impair mobility. Age brackets should also be accounted for, because elderly are more likely to have a slower pace than those in younger age brackets.
69
5.1.5 Video Surveillance Counting Techniques Further fieldwork should be done with video surveillance. More video clips recorded of more traffic situations should be counted. Traffic situations to count would be during heavy or light rainfall, dense or light fog, or festivals. Attempts should be made to obtain surveillance footage from the vigili urbani and the polizi locali to better determine the feasibility of counting from surveillance technology currently implemented in the city. More extensive research in recognition software should be carried out. An autonomous agent-based computer model to eliminate the reliance on man-hours is the ultimate end goal, and efficient software to complement the computer model would achieve that goal.
5.2 COMPUTER MODEL RECOMMENDATIONS The long term scope of the Venice Project Centerâ&#x20AC;&#x2122;s Mobility Teams has always been to address the mobility issues within Venice and to work towards improving mobility within the city. With this end goal in mind, the Mobility Team believes that the development of a fully functional computer model for the entire city of Venice will produce this desired result. 5.2.1 A Model Solution
A fully developed model has the potential to serve as a resource for government agencies and individual citizens. The prediction capabilities of a model will enable the City of Venice to take preventive rather than reactive measures to combat overcrowding and extreme traffic volume scenarios. The prediction capabilities as well as the ability of the model to fill in data gaps could allow Venetians to reroute themselves to avoid areas of high congestion, thus improving quality of life and helping to alleviate further congestion. 5.2.2 Collecting Data
Currently efforts to collect pedestrian traffic data are sporadic and expensive. Collecting data through field counts and interviews is very time consuming and leaves a lot of gaps in data. The Mobility Team recommends that instead of continuing to put money into surveys and traffic studies, a software counting system is developed.
Such a system would have the ability to conduct
pedestrian counts from video surveillance feeds and would be able to do so with minimal human input. This data could then be fed continually into the model as collected in order to increase the accuracy of traffic predictions. The use of such a system is certain locations would render the need for other pedestrian traffic data collection unnecessary. For instance if the video counting system 70
was implemented outside traghetto crossings the need to conduct manual counts or surveys at those locations would be no more. 5.2.3 A Comprehensive Network
The creation of pedestrian counting software could also be paired with existing video surveillance hardware already in place throughout the city. A number of city controlled organizations and agencies already have camera systems in place for various purposes. All of these video feeds if collected and coupled with the counting software could be dual-tasked to provide valuable traffic data that could improve the model and as a result mobility. Despite the multitude of cameras already in place, it is recommended that additional cameras be placed by the Department of Mobility at other key locations to provide a comprehensive picture of mobility in the city.
5.3 SMART-PHONE APPLICATION RECOMMENDATIONS
71
Bibliography Amilcar, Marcus, Amy Bourgeois, Savonne Setalsingh, and Matthew Tassinari. Mobility in the Floating City: A Study of Pedestrian and Water Transportation. Interactive Qualifying Project Report, Worcester: Worcester Polytechnic Institute: Venice Project Center, 2009. Black, Kevin, Sara Migdal, Michael Morin, Dukens Rene, and Nick Vitello. Urban Maintenance and Venetian Accessibility. Interactive Qualifying Project Report, Worcester: Venice Project Center, 2008. Bloisi, D., L. Iocchi, P. Remagnino, and N. Monekosso. "ARGOS-A Video Surveillance System for Boat Traffic Monitoring in Venice." International Journal of Pattern Recognition and Artificial Intelligence, 2009: 1407-1502. Carnevale di Venezia 2012. 2009. http://www.carnevale.venezia.it/ (accessed September 18, 2011). Catanese, Chris, Danise Chou, Bethany Lagrant, and Rudy Pinkham. Floating Around Venice: Developing Mobility Management Tools and Methodologies in Venice. Interactive Qualifying Project Final Report, Worcester: Worcester Polytechnic Institute: Venice Project Center, 2008. Catanese, Christopher D., Danice Yequay Chou, Bethany J. Lagrant, and Rudy E. Pinkham. "Floating Around Venice: Developing Mobility Management Tool." 2008. Centre, Flight. "Italy: A Unique Work of Art." Footprints, 2010. Cessi, Roberto, and John Foot. "Venice." Britannica Academic Edition. 2011. http://www.britannica.com/EBchecked/topic/625298/Venice/24379/Canal-boats-and-bridges (accessed September 16, 2011). Cessi, Roberto, Denis Cosgrove, and John Foot. Italy. 2011. http://www.britannica.com/EBchecked/topic/625298/Venice (accessed September 16, 2011). Chiu, David, Anand Jagannath, and Emily Nodine. "The moto ondoso index: Assessing the effects of boat traffic in the canals of Venice." 2002. Contesso, Lia. Venice Bridges. 2011. (accessed September 19, 2011). Davis, Robert C., and Garry R. Marvin. Venice, the Tourist Maze: A Cultural Critique of the World's Most Touristed City. Berkeley: University of California Press, 2004. 72
Drake, Cathryn. "Venice crossings: A traghetto tour reveals the city's other side." Wall Street Journal, 2008: sec World News. Duffy, J. "Re-Engineering the City of Veniceâ&#x20AC;&#x2122;s Cargo System for the Consorzio Trasportatori Veneziani Riuniti." 2001. Fiorin, Franco, and Giorgio Miani. "Development Plans for Urban Public Transport." Edited by Rinio Bruttomesso and Marta Moretti. Cities on water and transport, 1995: 100-107. How Stuff Works: Geography of Venice. 2011. http://geography.howstuffworks.com/europe/geography-of-venice.htm (accessed September 18, 2011). How Were Houses in Ancient Venice Designed and Why? http://answers.yourdictionary.com/entertainment-arts/architecture/how-were-houses-in-ancientvenice-designed-and-why.html (accessed September 7, 2011). Howard, Deborah, and S Quill. The Architectural History of Venice. Singapore: B.T. Batsford, Ltd., 2002. Italy. https://www.cia.gov/library/publications/the-world-factbook/geos/it.html (accessed September 15, 2011). Lopez, Angela. "Assessment of the Measure to Ease Pedestrian Congestion." Association for European Transport, 2006. Morgan, L. H. "The City of the Sea." Harper's New Monthly Magazine, 1782: 481. Ortalli, Gherardo, and Giovanni Scarabello. A Short History of Venice. Pisa: Pacini, 1999. Rameiri, E., V. Cogo, Mattei, and F.E.E. "Indicators of Sustainable Development for the City and the Lagoon of Venice." (Fondazione Eni Enrico Mattei) 1998. Riganti, Patrizia, and Peter Nijkamp. "Congestion in Popular Tourist Areas: A Multi-Attribute Experimental Choice Analysis of Willingness-to-Wait in Amsterdam." Tourism Economics, 2008: 1-15. Traffic Congestion Factoids. March 2009. http://www.fhwa.dot.gov/congestion/factoids.html (accessed September 14, 2011). 73
Van der Borg, Jan. "Tourism and Urban Development: The Case of Venice, Italy." Tourism Recreation Research 17, no. 2 (1992): 46-56. Van der Borg, Jan, and A. P. Russo. "Towards Sustainable Tourism in Venice." Sustainable Venice: Suggestions for the Future, 2001: 159-193. VeniceTable: Interactive Traffic Simulation Table. 2010. http://redfish.com/SFComplex/projects/veniceTable.html (accessed September 5, 2011). Zanini, Francesco, Fabio Lando, and Manuel Bellio. "Effects of Tourism on Venice: Commercial Changes over 30 Years." Working Papers, 2008.
74