Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
www.seipub.org/updr
A Dynamic Simulation of Crowd Flow in Taipei Railway and MRT Station by Multi-Agent Simulation System Kuochung Wen Department of Architecture and Urban Design, Chinese Culture University 55, Hwa Kang Road, Yangmingshan, Taipei 111, Taiwan wenkc@faculty.pccu.edu.tw Abstract Crowded together in Taipei Station, Taipei City, there are a number of flow transport behaviors, such as high-speed rail, rail and MRT transport. Timing and location require an established pattern in order for transport to flow without collision. Is there enough space for improvement in the form of moving lines to increase the flow of people? How can we reduce the possibility of disaster? These are important issues in the urban environment. A multi-agent system (MAS) is a computerized system composed of multiple interacting intelligent agents within an environment. MAS can be used to solve problems that are difficult or impossible for an individual agent or a monolithic system to solve. In this study, the establishment of MAS architecture assessment of a flow dynamic simulation method for dynamic simulation of combined time and patterns of behavior analysis to act as a space path performance assessment and inspection certificate is one effective tool. Keywords Multi-Agent Simulation System; Crowd Flow; Moving Path; Dynamic Simulation; Simulation-based Evaluation
Introduction Crowded together in Taipei Station, Taipei City, there are a number of flow transport behaviors, such as the Taiwan High-Speed Rail (HSR), the Taiwan Railway (TRA) and the Taipei MRT. A daily average of 30 million passengers in transit during rush hour and the MRT train classes varying from regulation, coupled with the four underground moving lines, creates a complex network. The goal is for these systems to flow past each other, avoiding collision and crowding. Timing and location have an established pattern to follow. How is it possible to increase the flow of people and reduce the possibility of disaster? These are important issues in the urban environment. A multi-agent system (MAS) is a computerized system composed of multiple interacting intelligent agents within an environment. MAS can be used to solve
problems that are difficult or impossible for an individual agent or a monolithic system to solve (Almeida et al. 2013). The people in the MRT of Taipei Station are independent but interactive. However, with so many people in each person’s way, the paths, speed of walking, cognition of the place, sight lines and the interactive characteristics of behavior all differ. This creates a problem in how to simulate every individual for analysis and assessment. In this study, the establishment of MAS architecture assessment of the flow dynamic simulation method’s combined time and patterns for behavior analysis acts as path performance is an effective tool. Theoretical studies include MAS, the geographic information system (GIS), spatial statistical analysis and dynamic simulation methods. Since the crowd has time-varying and non-linear flow characteristics, etc., the phenomenon has a very complex spatial behavior in the simulation. Also, the spatial environment for the overall model of institution buildings and behavioral activities related to import and spatial information content calculations, must be evaluated with the crowds’ cycle and behavioral patterns related to the analysis before the final establishment of crowd flow patterns for the dynamic simulation system for authentication. Reviews Crowd Simulation As crowd behavior simulation has been an important goal of computer animation; there is considerable research in motion simulation, path planning and navigation, including acts of large-scale virtual reality simulation environment and cognitive patterns in both directions. However, the mass flow of the study consists of three main models: social dynamics and cellular automata, inference rule. Despite considerable 59
www.seipub.org/updr
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
efforts devoted to each to create realistic simulation results, there is no single model that can improve the mass density. Social dynamics simulations, used to establish the flow of particles, tend to look like the movement of people. For example, the cellular automata model for movement and space is limited at high density because inference rules of the model cannot control high scale panic situations (Oguz, 2010; Zhou, 2010; Paris 2009; Banerjee 2009; Bandini 2007; Pelechano 2006). However, HiDAC (High-Density Autonomous Crowds) is capable to solve a portion of the above problems. The application of multiple agents can be a more natural way in the dynamic virtual environment to simulate the flow of a high-density population (Morillo 2004; Pelechano 2006). A solution to the problem is the actual simulation of the local movement under different circumstances, with different agents of personality psychology, including physiological and geometric rules combined with physical strength. Basically, all agents use the same rules of operation, resulting in a uniform act, but giving an agent different psychological attributes (such as impatience, panic, personality traits, etc), physiological attributes (such as exercise, energy levels) and other personal characteristics, which triggers different behaviors. Each agent behaves differently when given static and dynamic objects and the space in the vicinity of other agents’ perceptions and responses (Kolli et al., 2013; Araujo et al., 2013). Multi-agent Simulation System (MAS) In this study, the MAS theory used has analogy system simulation concepts and a theoretical basis. The main actors will be the basis of mode and grid mode combined to produce dynamic simulation scenarios. Multi-agent systems can be defined in a common environmental group of actors in interaction; and actors can fix themselves and their environment. MAS extends the research tools and results, from the simplest Boids Model, through the Santa Fe Institute, Massachusetts Institute of Technology (MIT) Media Lab, London Space Analysis Center (The Centre for Advanced Spatial Analysis, CASA); other colleges continue to develop software with relatively simple operation and programming languages to help non-worker applications in various fields (Si-wei, 2006). A few are as follows: 1. Biological cluster of the Boids Model (Reynolds 1986). 60
2. Application of a parallel update of the concept of constructing cellular automata (Cellular Automata, CA). 3.
Cognitive behavioral interactions, the SWARM and MAML.
4. Simulation of natural phenomena and the formation of Starlogo learning and NetLogo. 5.
Integrated regional transportation system analysis, environment Transportation Analysis Simulation System (TRANSIMS).
6. Simulation of different levels of pedestrian movement, STREETS (Haklay, 2001). NetLogo NetLogo is a natural and social phenomenon used to simulate the design of a programmable modeling environment. Uri Wilensky in 1999, initiated by the connected learning and computer-modeling center (CCL), was responsible for continued development. It was particularly suitable for the time evolution of complex systems modeling. Modelers can run hundreds of separate agents to issue instructions. This allows the exploration of the micro level of individual behavior; and the macro mode of possible links between these macro-models is composed of many individual interactions emerging between them. The simulation model covers the natural and social sciences in many fields, including: biology and medicine, physics and chemistry, mathematics and computer science and economics and social psychology. The NetLogo platform has been widely used by the majority of users (Sklar 2007); the model can be used as Java Applets running in a browser, an operating system from the Interface, Procedures and Information Form. The NetLogo model is divided into three parts: turtles, on behalf of those behaviors; patches, from all environments composed of all the grids, so that the patches can be observed on behalf of our environment; and observer, so that what we call life or world view, can represent the environment or mode the user observed (Batty, 1999). Procedures and variable declarations are made by some of the components. The analogy system is created by what the patch(es) and turtle(s) posed. Only if the system automatically gives a turtle numbers, can patch and turtle be used for individual properties or change its behavior (Lin, 2008). Studies were able to construct MAS land use changes along the MRT analogy mode and change control information in previous years with more than 90
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
percent accuracy (Zhang, 2006; Yan, 2005). Another study discussed the building's contours and visitors’ fixed-line analogy, obtaining good interpretation of the capacity (Si-wei, 2006), to explore the impact of spatial cognition choice behavior to establish the spatial relationship between reasonable and appropriate for the network. A very high linear correlation was found in the interaction between space and pedestrians, based on the theory of spatial cognition, the use of sight and the sight of the two properties as spatial cognition (Chen, 2004). Another square of the pedestrian movement characteristics used to build a NetLogo simulation model, to analyze the behavior of pedestrians' movements, shows a good interpretation of the simulation model; in addition, motion simulation can be further applied to the population simulation (Yang, Liu, Li, Zhu & Fang, 2009).
www.seipub.org/updr
data and spatial data conversion, dynamic simulation and evaluation basis for the application of the data structure for the behavior of the entire crowd, for the third part of the simulation and evaluation module input information purposes. The third part, a simulation evaluation module, is divided into the crowd dynamic module, analysis module, the line planning evaluation module, the potential risk assessment module and the improvement policy evaluation module. The dodge mode is determined by the flow of people, the knowledge base for the core of the whole operation, the road mode, accumulation mode and other factors entering the settings, and then output through the system interface GUI with a 3D simulation.
Theory Conception Model The overall structure of the entire study contains the following five sections, as shown in Fig. 1. The first part is the underlying database, database management system, as well as the basic database of the survey analysis and implementation, including: attribute information (such as spatial attributes, path attributes, flow property, facilities and equipment, building codes, text statistics or numbers) and space (layout diagram of the path network diagram, equipment distribution, crowd moving path information).
FIG. 1: RESEARCH METHOD ARCHITECTURE
The second part includes: the GIS processing platform, the main attribute conversion, the raster data analysis operations to the associated processing and attribute
FIG. 2: DYNAMIC SIMULATION OF MULTIPLE LAYERS OF FIRE ESCAPE STACKING CONCEPT MAP
The fourth part is the computing platform of the multi-agent simulation system, the use of the NetLogo programming setting crowd flow behavior, environmental interaction factors, the crowd dynamic simulation and crowd interactive mode, given the different characteristics of different agents, ability and goals in the space environment fire escape via the API through interaction, dynamic simulation and evaluation. In this study, the concept of a multi-layer stack is used to classify the space environment information in order to clearly express the structure of the overall relationship with the individual. Therefore, it is helpful not only for the establishment of property information and a display object, but also in terms of carrying out operations on the NetLogo simulation, so that it is easy to operate and analyze logic functions. These are divided into space-based environmental information (like plan, zone, moving path, factors, etc.) and crowd behavioral information (like walking, dodging, gathering, etc.). Then, they are assessed for decision-making and to evaluate the effect of the crowd results for the four categories shown in Fig. 2. 61
www.seipub.org/updr
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
Cellular Automata Simulation of Behavior Field Model The Cellular Automata (CA) theory originated is in the early 1940s. It was the concept used by most physicists Ulam mentioned, but later von Neumann (1966) used this concept to study the characteristics of the self-replicating system of logic, and in 1970 a British mathematician (Conway, 1970) invented the "game of life" to further develop mathematical models that could be used (Wolfram, 2002). In this study, the dynamic selection process was used to carry out a dynamic simulation of a fire escape. Mathematical equations expressing the individual or activity patterns are generated by a certain inference (Wolfram, 1984; 1994). Using the escape path of evolution from the concept of an asylum, (Conway, 1970) invented the "game of life" as the basis for action, and then expressed this knowledge, including cell description, rules and the multi-cell stack overlap and so on. Multi-layers were overlapped by different generations of a series of a process, not only for different attributes of the dynamic visual simulation of planar objects, but also by different time series to show different layers of information, such as disasters, the spread, the best escape routes, and so on, that could provide a reference. These only showed the effect of the plane, and belonged to the integrity of the dynamic assessment (Lau, 2002). 1) Spatial Scale in Multiple Overlapping Conducting the operation in this part is different because the object has different scales, corresponding manners and, according to this study, different scales of space factor and, to overlay, select a common reference point in space by the various impact factors in different space scales in order to overlay and show the contents of the overlay status, as shown in Fig. 3.
FIG. 3: SCHEMATIC DIAGRAM OF OVERLAPPING SETS OF DIFFERENT SPATIAL SCALES
2) Time Scale in Multiple Overlapping This iterative process of the basic plan in the time series during the unit time may not directly correspond to the same point on the timeline, so this study takes this point of time T2 in Fig. 4, to establish the Ta and Tb time-points status of
62
possible related factors. The time axis corresponds to T, and then T1 to T2 in order to carry out different time units of the basic map overlay.
FIG. 4: ITERATION TIME IN THE OVERLAYING OF DIFFERENT TIMING DIAGRAM
Multi-agent Simulation System The NetLogo system program includes the "interface", "procedures" and "information". The NetLogo model is divided into three parts: turtles for behavior; patches from all environments composed of all of the Palace box, so patches can be observed on behalf of our environment; and the observer we call life, which has a world view, and which can act on behalf of the environment or mode the user observes (Batty, 1999; Lin, 2008). Turtles are autonomous actors who exist in a patch formed by the grid of cell space; and each cell is a grid patch. Actors, in addition to their own characteristics, are affected by other actors when making a decision. Between the patches: the path is from the perspective of asylum; the patch between the role of cellular automata will be produced; and each patch will be affected by the neighboring patch. Between turtles and patch: the actors themselves will be environmental impacts. There are different decision-making behaviors; and relative environmental decisionmaking will happen because the actors change. They will affect each other. We can use various factors in the changes in composition to determine whether or not in the environment. Between the observer and each element: the observer in the model actually refers to the operator. We must set some things that cannot be changed and research limitations. In this study, the overall operational concept is to create walking impact factors equal to individual grid layers (patches) in the operation. People’s behavioral responses are similar to turtles’ agents. In addition, the various layers will be stacked with this in mind, but also keep the other layers mutually correlated. In the iterative process which is the time of reaction, the process also added speed and distance factors, and a complete simulation of the interaction of time and space systems. Therefore, this study proposed not only a single dimension (plane grid) of evolution, but also
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
multiple, interactive dimensions (plane grid with different themes) for a simple to complex simulation process. This is the most central part of this study. Different types of simulation to assess time and space will produce different changes, so we can develop their different themes for different strategies and check the effectiveness of its prevention. Field Model In the mass flow behavior during simulation, we must consider the social psychology of the important elements of the action to set the rules for the simulation. In the case of a calm setting, people will go directly along the shortest path to the exits. However, emotional confusion easily influences others, and changes their actions, so that one person blindly follows another. In the domain model, the judgment should follow along the shortest path to escape the action of others or the extent of crowd evacuation action, and is considered the most important factor. The ratio between the two sets a Panic Level. The greater the panic, the stronger the urge to follow others! If we want to follow others we will deviate from the shortest path to the exits. Consider the domain model in regard to other people's property. We assume the footsteps of people will stay on the ground, and are more likely to follow others, which moves the footprint to more places. The amount of information in this footprint is called the Dynamic Field; the shortest distance information is Static Field. A footprint based on the number of changes at each time point is dynamic. The shortest distance does not change, it is static. Field simulation gives light to these two games, to take refuge in the mode of action. According to the static field or the field that decides where to move one cell forward, the decision is based on probability. The higher the level, the more likely the panic! In the mobile lattice footprint, the lower the level of panic, the more likely it is for people to follow the shortest path (Jutai, 2008). In Fig. 5 digital values represent the Field (distance/footprint). The left display chooses the shortest path direction; the right displays distance to the tracks depending upon direction.
â—?
50/04 â—? 52/02
â—? 51/10 â—?
â—?
50/04 â—? 52/02
â—? 51/10 â—?
FIG. 5: (LEFT) FROM / FOOTPRINT (RIGHT) FIELD VALUE OF FORWARD OPTIONS (JUTAI, 2008)
www.seipub.org/updr
Globe and Local Field This study also sets the crowd type properties and selects the direction of the related field's role. Basically, it can be divided into Global and Local fields, such as building components (walls, columns, doors, windows, etc.) A field is part of a global field; and all types of people are subject to its role. Other examples are footprints, indicators, weights, etc. The field is a part of the regional field. Only some of the relevant population can use its role type, such as a footprint, if they have a role to follow. However, the field is determined from the weight of a particular space for specific populations, and the formation of an attachment, such as the space inside the distance, is weighted to leave room for those who only have a role. An exit door for most of the refuge will be useful; those panicking or who cannot get enough information about the role of space are greatly reduced. Another concept is the sight range. The sight range of different groups should differ according to their attributes, so sight does not refer to really seeing the sights but a sense of space awareness. Sometimes work experience relevant to the mandate may be related to its properties. In general, adults and children have better sight than the elderly. Security and staff can identify their sights for the whole visual field, regarded as the MRT station. On the whole, for those who just come to the MRT station, it is probably the only sight is the space around their place, but indicators such as the provision of spatial information will increase the range of sight. Flow and Density Flow simulation mainly uses flow (Q) and the pedestrian density (K) to describe these two parameters in which the flow and density are relevant. Flow, per unit of time, through a number of points or sections of lines, is generally 1 min or 15 min. Therefore, in human traffic units/min, the unit width flow refers to the effective width of pedestrian traffic through the unit, the unit man/(m * min). In this study, simulations can be a direct output per unit of time for the number of people, and in some areas we may be able to estimate the traffic flow changes by the regional distribution of pedestrian movement. Space per unit time flow rate = the number of changes: P (n +1) −P n
đ?‘„đ?‘„ = T
(n +1) −T n
Equation 1
where Q is the flow, P is the population and T is time (ticks). Pedestrian density is per unit area or study area,
63
www.seipub.org/updr
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
walkway or pedestrian queuing area in the average number of units of human/m2. In a sparse crowd, pedestrians can choose their preferred velocity, but in crowded conditions, the flow and speed will decrease. With a positive stimulus, a pedestrian under normal circumstances can accept a higher density. In this study, we calculate the spatial density of the simulation system within the grid by the number of analogy systems and spatial statistics to obtain the number of the two values of flow and density changes observed in the simulation system of relations: P
đ??žđ??ž = A
Equation 2
where K is the density, P is the population and A is the number of room’s grids. Simulation Simulation Environment Parameter Setting
that users can intuitively observe the overall simulation scenarios, it shows zone blocks, paths and personnel walking behavior developments. (2) Walking simulation: Dynamic and server personnel move by the path near the grid within a certain radius. Any action is a departure from the minimum in the direction; its speed is according to an initial set of types; both floors run down and out. Numerical Simulation Analysis and Evaluation Numerical simulation analysis and evaluation include: (1) The total population assessment: It includes the total number of escapes, the death toll, the total number of people who fled outdoors, the total number of injuries, the total number of people trapped and security.
Simulation environment parameter settings include:
(2) Bottleneck area:
(1) Data reading:
(3) The state is marked by different directions in an attempt to find how different patterns may affect walking. Assessment of the floor population retention:
This includes city planning of street blocks and building blocks. Property settings include: the building floor, building projected areas, building numbers, building construction, street block numbers and street blocks of land use types. (2) Setting the scene: A population view of the setting of the system is limited to the path and speed, including the random set with its own set of two, in which the random set can also adjust its density. (3) Staff setting: Basically it is a place only for walking personnel, in order to avoid overloading the system operation, which can adjust its density. Staff configured by the system constitutes a random number, starting with street scenes, building floor, and so on. (4) Walking direction control: Walking direction is divided into four cardinal points and other combinations of eight directions, coupled with a windless situation; this is the leading direction. Dynamic Simulation of Iterative Evolution Dynamic simulation of iterative evolution includes: (1) Scene simulation: A provider of a visual output screen window so 64
Records are subject to escapes made by people on each floor of the building in a time series of changes, in an attempt to find possible models. (4) The path zone in assessing: The number of people, zone blocks, paths, stair blocks and other zones will be explored for the various zone blocks for path and space characteristics under the impact of personnel crowding. (5) The street blocks for crowded areas: Record every time the crowded area of the zone is blocked to explore the various zones for blockage and space characteristics showing the impact of density. Simulation Place The Taipei MRT systems in Taipei MRT stations are the main research sites. The main station, according to the system’s point of view, is divided into the TRA, HSR and MRT mass transit systems, consisting of three parts: terms of the building structure, six stories above ground and an underground four-story building, as shown in Fig. 6. The study observes the ground floor main hall, the TRA, the ticket booth and HSR passenger seating area; the TRA, the HSR
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
platform layer on the ground and second floors; the MRT Bannan Line (blue line) and the Danshui line (red line) monthly units are located underground (third, fourth). From above the spatial distribution can be observed. The HSR, TRA and rapid transit routes and mobile platforms are almost all concentrated from the basement to the ground floor fourth. The sight of such a complex environment will make it difficult for the user to find correct directions. This study considered the importance of physical space in the Taipei rail station system, the space complexity of the district and the Taipei MRT system in Taipei station before choosing the MRT for its study sites. In this study, passengers used public space. Exploring the engine room, office and other employees could not be covered in the study range.
www.seipub.org/updr
space used to do the segmentation and analysis of the project. First, the general perception of space in accordance with the type of plane is divided into the entrance hall, platforms, stairs, escalators, gates, fire doors and compartments. The empirical research showed all of the space at the MRT Taipei Main Station, (for example, a total of four floors, eight general MRT entrances and exits, eight entrances and exits to the TRA station, six HSR station entrances and exits, eight entrances, staircases and escalators to the business district and 15 points of other vertical movement, two platforms and eight halls). The moving target was based on the type of space and property to be subdivided into general moving space (zone), vertical movement space (stair), exit/entrance space (mrt), entry and exit gate (gate), commercial entrances (shop), railway system entrances (trans), station entrances and exits (out), fire doors (fgate) and escape doors (exit). GIS collates those properties with their establishment in Fig. 7. TABLE 1: PROPORTION OF THE TAIPEI MRT HOURLY TRAFFIC
FIG. 6: OVERVIEW MAP OF TAIPEI STATION
Time Traffic Volume (%) Hour Hour Weekday Holiday 15-16 6-7 1.92% 1.08% 16-17 7-8* 7.83% 2.88% 17-18* 8-9* 11.97% 4.51% 18-19* 9-10 5.52% 4.76% 19-20 10-11 3.65% 4.91% 20-21 11-12 3.11% 5.27% 21-22 12-13 3.72% 6.56% 22-23 13-14 3.95% 7.28% 23-24 14-15 3.72% 6.86% 24-01 * Rush hour (Source: MRT Corporation)
Weekday 3.75% 4.93% 8.50% 11.39% 7.63% 5.04% 5.55% 5.50% 1.84% 0.48%
Holiday 6.83% 7.20% 8.58% 8.02% 6.20% 5.23% 5.59% 5.24% 2.27% 0.72%
Time and Population Setting
FIG. 7: TAIPEI MAIN STATION SIMULATOR
Simulator To consider the operation of the computer system performance and computing time required for analysis, we used the grid scale of 1.5 m, in order to achieve efficiency and accuracy of the simulation and create a balance. The simulation grid type classification system is based on the type of rapid transit stations within the
For the MRT Taipei Station at rush hour, the average class distance is about 6 min, with an overlapping interval of about 3 min during off-peak periods. The average class distance is about 8 min with about 4 min interval overlaps. At 23:00 after the shift, it widens to 12 min with overlapping intervals of 6 min. Simulation systems must correspond to real-time. Tick the NetLogo simulated counter to set it to a certain number of times and the system will stop the simulation. You must use the same simulation time as a benchmark to compare the simulations to the differences. In this study, each tick is considered a real time of 1 sec, corresponding to the actual time of 60 ticks 1 min, which is given to simulate the performance limits of computing devices. For the sake of efficiency required for the actual time of the tick ratio adjustment, it is also easy to follow and compare the actual transit flow verification. Adjusting the 65
www.seipub.org/updr
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
number and spacing of stops is according to the MRT Taipei Station for setting off-peak hours. To operate conditions for the full day simulation, the simulation models should be 68,400 ticks of the simulation, as in Table 1.
TABLE 3: THE EVALUATION OF FLOW CROWD STATISTICS
1. Transit performance evaluation: the flow of population moving by every transit pattern. M1m3 is from Platform m1 to Platform m3. The max is about 223 people per tick. The daily total of people is 300,000.
TABLE 2: THE FLOW CROWD STATISTICS
Ticks=14500
Ticks=37700
Ticks=49300
2. The total number of statistics on each floor: the flow of population moving on every transit floor. B1 is basement 1. The max is about 4,460 people per tick at B1.
3. Traffic statistics at each gate, gate efficiency: the flow of population walking through the ticket gate. G11 is the number of gates. The max is about 13 people per tick at g21.
4. The point flow statistics of vertical movement: the efficiency of mobile stairs. The flow of population moving on the stairs. S1 is the number of stairs. The max is about 164 people per tick at s8.
5. Statistics on the number of zone blocks: the spatial density calculation. The flow of population moving in every zone block. The max is about 3,033 people per tick at z3.
Simulation Results To observe changes in the flow space and the overall transport system efficiency, the number of the simulation system is calculated as the number of transfers, the total number of statistics on each floor, each vertical set of stairs, each gate and each block. The main function of these calculations includes: 1. Transporter demographics: transit performance evaluation. 2. The total number of statistics on each floor: space and size distribution. 3. Traffic statistics at each gate: gate efficiency. 4. The point flow statistics of vertical movement: the efficiency of mobile stairs. 5. Statistics on the number of zone blocks: the spatial density calculations are shown in Tables 2 and 3. 66
Conclusions In this research, we created a crowd flow simulation model, a comparative analysis of different patterns of transit space, and flow data, as well analysis of the overall moving productivity and effects was conducted. In the present study, it was found that pedestrians’ basis for spatial perception is the Static Field used to plan the shortest path and vulnerability to Dynamic Field effects of a moving line. Because of these complex features, pedestrians and the moving
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
www.seipub.org/updr
line and zone interact with good conditions and evolve over time, gradually forming a fixed pattern.
conference on Autonomous agents and multi-agent
In this study, the relationship was observed between space density and time delay of a crowd. That is, overall efficiency increases as the density and delay variability also increases. A reduction in the flow path shows that a high density can cause system instability, but to some extent after the move, it can form a stable state.
for Autonomous Agents and Multiagent Systems.
Subsequent use of high-flow analysis to find congestion positions and modify the paths with partitions (the current MRT system often uses the body's moving-lane operation) was found to effectively improve overall efficiency and obtain an optimal solution that can be improved in other multiple simulation tests. Because of the time and hardware constraints, the present study only conducted an in-depth analysis of the MRT Taipei Main Station with the moving path in the zone space alone. In a follow-up study, it was proposed to observe other types of rapid transit stations and simulations to compare the flow characteristics of the same types of station, to establish general rules. In this study, especially in walking space environment observations for vertical moving lines, the export of all types of transport zones influences the flow of people, but taking into consideration issues such as time and data collection, the commercial interior space for more stations was ignored. In a follow-up study, it was hoped to include more details regarding MRT stations and commercial space analysis of its attractiveness to transit passengers. A lack of rapid transit systems for more statistical data makes the adjustment of simulated data in this study not precise enough. In the future, if such detailed data are obtained, it is possible to adjust in and out of stations more accurately, so that the overall simulation is more realistic.
J.E.,
Simulation
Bandini, S., M. L. Federici, and G. Vizzari. "Situated Cellular Agents Approach to Crowd Modeling and Simulation." [In English]. Cybernetics and Systems 38, no. 7 (2007): 729-53. Banerjee, B., A. Abukmail, and L. Kraemer. "Layered Intelligence for Agent-Based Crowd Simulation." [In English]. Simulation-Transactions of the Society for Modeling and Simulation International 85, no. 10 (Oct 2009): 621-33. Batty, M., Dodge, M. and Jiang, B. Gis and Urban Design. Geographical Information and Planning. edited by F. Stillwell, and Geertman, S. Berlin: Springer-Verlag, 1999. Chen, Zhengde. "Spatial Structure and Pedestrian Movement Simulation - a Case Study Simon Foot. Agent-Based Modeling of Spatial Configuration and Pedestrian Movement." National Taiwan University, 2004. Conway, John Horton. "The Game of Life." Scientific American 223, no. 4 (1970): 4. Haklay, M., D. O'Sullivan, M. Thurstain-Goodwin, and T. Schelhorn. ""So Go Downtown": Simulating Pedestrian Movement in Town Centres." [In English]. Environment and Planning B-Planning & Design 28, no. 3 (May 2001): 343-59. Jutai, Gaku. The Crowded - Person, Car, Ants, Internet, Cell All the Way through the Secret [in Chinese]. Taipei: Whether, 2008. Kolli,
S.,
&
Karlapalem,
K..
“MAMA:
multi-agent
management of crowds to avoid stampedes in long queues.” In Proceedings of the 2013 international conference on Autonomous agents and multi-agent systems (May 2013): 1203-1204. International Foundation for Autonomous Agents and Multiagent Systems.
REFERENCES
Almeida,
systems (May 2013): 1351-1352. International Foundation
Kubera, Y., Mathieu, P., Picault, S., "Everything can be Rosseti, Modeling
R.J.F.,
Coelho,
Applied
to
A.L.,
“Crowd
Emergency
and
Agent!" (PDF), Proceedings of the ninth International Joint
Conference
on
Autonomous
Agents
and
Evacuation Simulations using Multi-Agent Systems” -
Multi-Agent Systems (AAMAS'2010) (Toronto, Canada)
arXiv preprint arXiv: 1303.4692, 2013 - arxiv.org
(2010): 1547–1548.
Araujo, F., Valente, J., Al-Zinati, M., Kuiper, D., &
Lau, K. H. "A Cellular Automation Model for Urban Use
Zalila-Wenkstern, R., “DIVAs 4.0: a framework for the
Simulation." Paper presented at the Proceeding of 25th
development of situated multi-agent based simulation
Australasian Transport Research Forum Incorporating
systems.” In Proceedings of the 2013 international
the BTRE Transport Policy Colloquium., 2002.
67
www.seipub.org/updr
Urban Planning and Design Research (UPDR) Volume 1 Issue 4, December 2013
Lin, F. T. http://bp119.bp.ntu.edu.tw/netlogo/ netlogoprogs. htm.
Song, Si-wei. "Taroko National Park Visitor Center Visitor Behavior." National Taiwan University, 2006.
Linse, T., and J. Natterer. "A Seven Floor Residential
von Neumann, J. In Theory of Self-Reproducing Automata.
Building Almost Completely Made of Timber - Report of
Fourth University of Illinois Lecture. edited by A. W.
a Pilot Project." Bauingenieur 83 (Dec 2008): 531-39.
Burks Urbana: University of Illinois, 1966.
Morillo, P., M. Fernandez, and N. Pelechano. "A Grid Representation for Distributed Virtual Environments." [In English]. Grid Computing 2970 (2004): 182-89. Oguz, O., A. Akaydin, T. Yilmaz, and U. Gudukbay. "Emergency
Crowd
Simulation
for
Outdoor
Environments." [In English]. Computers & Graphics-Uk 34, no. 2 (Apr 2010): 136-44. Paris, S., and S. Donikian. "Activity-Driven Populace: A Cognitive Approach to Crowd Simulation." [In English]. IEEE Computer Graphics and Applications 29, no. 4 (Jul-Aug 2009): 34-43. Pelechano, N., and N. I. Badler. "Modeling Crowd and Trained Leader Behavior During Building Evacuation." [In English]. IEEE Computer Graphics and Applications 26, no. 6 (Nov-Dec 2006): 80-86. Reynolds, V. "Biology and Race-Relations." [In English]. Ethnic and Racial Studies 9, no. 3 (Jul 1986): 373-81. Sklar, E. "Software Review: Netlogo, a Multi-Agent Simulation Environment." [In English]. Artificial Life 13, no. 3 (Sum 2007): 303-11.
68
Wolfram, S. A New Kind of Science.
Champaign ,Illinois:
Wolfram Media, Inc., 2002. Wolfram, S. "Cellular Automata as Models of Complexity." Nature 311 (1984): 6. Yan, Ziyang. "Rapid Transit Along the Land-Use Change Simulation
Model
Construction
and
Application."
National Chiao Tung University, 2005. Yang, L. Z., S. B. Liu, J. Li, K. J. Zhu, and T. Y. Fang. "Information-Based Evacuation Experiment and Its Cellular Automaton Simulation." International Journal of Modern Physics C 20, no. 10 (Oct 2009): 1583-96. Zhang, Di Zheng. "Mrt Station Simulation of Land Use Change - a Case Study of Taipei Mrt Mucha Line." National Chiao Tung University, 2006. Zhou, S. P., D. Chen, W. T. Cai, L. B. Luo, M. Y. H. Low, F. Tian, V. S. H. Tay, D. W. S. Ong, and B. D. Hamilton. "Crowd Modeling and Simulation Technologies." [In English]. Acm Transactions on Modeling and Computer Simulation 20, no. 4 (Oct 2010).