Produced by
Navigating the future of mobility What tasks can AI effectively and safely perform in the mobility sector? SEPTEMBER 2024
ITSInternational.com
INTRODUCTION
AUTHOR: Alex Nesic is a serial tech entrepreneur who has worked at the intersection of AI and urban mobility, most recently as a co-founder of Drover AI
A revolution in t Navigating the future of mobility means approaching AI as a powerful tool that, when wielded responsibly, can help us build transportation systems that truly serve people -
The transportation sector is undergoing a profound transformation, driven by the rapid advance of artificial intelligence (AI). As we drive towards an increasingly interconnected and data-driven future, AI is emerging as a gamechanging force in how we move people and goods across the globe. From bustling city centres to remote rural highways, AI is reshaping our mobility systems, promising unprecedented levels of efficiency, safety and sustainability. Yet AI remains an enigmatic concept, often shrouded in a mix of excitement and apprehension. While sexy tools like Chat GPT and Midjourney dominate the broader AI conversation, there are myriad more mundane yet powerful applications of AI in transportation. As we venture down this path, it’s crucial to demystify AI, understand its real-world applications, and critically examine both its immense potential and the challenges it presents.
-
-
In transportation, AI manifests in various forms, from machine learning algorithms that optimise traffic flow to computer vision systems that govern micromobility fleets or enhance vehicle safety. By understanding these fundamental concepts, we can better appreciate AI’s role in shaping the future of mobility.
Key applications of AI in transportation
What AI currently is and isn’t capable of
Traffic management and urban mobility AI is revolutionising how cities manage traffic and urban mobility. Intelligent traffic systems use real-time data from sensors, cameras and connected vehicles to dynamically adjust traffic signals, reducing congestion and improving traffic flow. AI algorithms can predict traffic patterns, allowing city planners
At its core, artificial intelligence refers to computer systems capable of performing tasks that typically require human intelligence. These include visual perception, speech recognition, decision-making, and language translation. However, it’s important to dispel some common misconceptions:
SEPTEMBER 2024
AI is not a magic pill that can instantly solve all problems. It’s a tool that requires careful implementation and human oversight. Current AI systems are “narrow” AI, designed for specific tasks. We’re still far from “general” AI that can match human-level intelligence across all domains. AI doesn’t (yet) “think” like humans. It uses complex algorithms and vast amounts of data to recognise patterns and make predictions.
2
ITSInternational.com
A I I N T R A N S P O RTAT I O N
n transportation Autonomous vehicles and micromobility Perhaps the most visible application of AI in transportation is in the development of autonomous vehicles (AVs). Self-driving cars, trucks, and buses use a combination of sensors, cameras and AI algorithms to navigate roads safely. While fully-autonomous vehicles are still in development, AI is already enhancing vehicle safety through advanced driver assistance systems (ADAS). In the micromobility sector, AI is optimising the deployment and management of shared bikes, scooters, and other small vehicles. AI algorithms can predict demand patterns, ensuring vehicles are available where and when they’re needed most. Computer vision technology is being used to enhance safety, for example, by detecting when riders are not wearing helmets or detecting what type of infrastructure (sidewalk, bike lane or street) is being used and sharing that data with urban planners.
to proactively manage transportation networks and respond to changing conditions. For example, some cities are implementing AI-powered traffic management systems or using AI-enabled crowd sourced dashcam video feeds that can optimise routes and reduce average travel times by up to 25% and cut emissions by adapting to realtime traffic conditions. These systems can also prioritise public transit and emergency vehicles, further enhancing urban mobility. Public transit optimisation AI is making public transportation more efficient, reliable, and user-friendly. Machine learning algorithms can analyse historical and real-time data to optimise bus and train schedules, predict maintenance needs, and improve route planning. AI-powered chatbots and mobile apps are enhancing the passenger experience by providing personalised trip planning and real-time updates. In some cities, AI systems are being used to dynamically adjust bus routes based on demand, reducing wait times and improving service in underserved areas. Computer vision systems are also put to use on buses to detect and enforce vehicular violations of dedicated bus lanes which helps ensure timely bus service. This not only increases ridership but also contributes to more equitable and accessible public transportation. SEPTEMBER 2024
Tolling and congestion charging AI is streamlining tolling and congestion charging systems, making them more efficient and fair. Machine learning algorithms can analyse traffic patterns to implement dynamic pricing, encouraging off-peak travel and reducing congestion. AI-powered image recognition systems can accurately identify vehicles and process payments without the need for physical toll booths, reducing traffic bottlenecks. Similarly, some of the most hotly-
3
ITSInternational.com
INTRODUCTION
AI can monitor driver behaviour and alertness, preventing accidents caused by fatigue or distraction. The potential impact is significant – some estimates suggest that widespread adoption of AI-driven safety technologies could reduce traffic fatalities by up to 90% in the long term.
contested real estate in cities is the kerb; AI-powered computer vision systems are being deployed to proactively manage and enforce proper use of it by various stakeholders depending on the time of day or week. Beyond that, these systems can be integrated with broader urban mobility strategies, using AI to optimise pricing and incentives to achieve specific policy goals, such as reducing emissions or promoting public transit use.
Environmental sustainability As the world grapples with the urgent need to address climate change, AI offers promising solutions for reducing the environmental impact of transportation. By optimising routes, reducing congestion and improving vehicle efficiency, AI can significantly cut fuel consumption and emissions. In public transit, AI-driven systems can encourage ridership by improving service reliability and convenience, potentially reducing private car usage. Moreover, AI is playing a crucial role in the integration of electric vehicles and renewable energy systems, helping to balance grid loads and maximise the use of clean energy in transportation.
Logistics and supply chain In the world of freight and logistics, AI is driving significant improvements in efficiency and sustainability. AI-powered route optimisation algorithms can reduce fuel consumption and delivery times. Predictive maintenance systems use machine learning to anticipate equipment failures, reducing downtime and extending the lifespan of vehicles. AI is also enhancing warehouse operations through automated inventory management and robotic picking systems. In ports and intermodal facilities, AI is being used to optimise container stacking and vessel loading, significantly reducing turnaround times.
Improved user experience AI is transforming the way people interact with transportation systems, making journeys more personalised, convenient, and enjoyable. From AI-powered trip planning apps that provide real-time, multimodal journey options to chatbots that offer instant customer support, the technology is putting users at the centre of the mobility ecosystem. In the future, AI could enable even more seamless and intuitive transportation experiences, such as autonomous vehicles that adapt to individual preferences or smart cities that anticipate and meet citizens’ mobility needs proactively.
Benefits and opportunities Increased efficiency and productivity AI’s ability to process vast amounts of data and make rapid decisions is driving unprecedented efficiency gains across the transportation sector. From optimizing traffic flow to streamlining logistics operations, AI is helping to reduce delays, cut costs and improve resource utilisation. For instance, AI-powered predictive maintenance can reduce vehicle downtime by up to 20%, while route optimisation algorithms can cut fuel consumption by 15% or more.
Challenges and considerations Data privacy and security The effectiveness of AI in transportation relies heavily on data, raising important questions about privacy and security. As vehicles and infrastructure become more connected, they also become potential targets for cyberattacks. Ensuring the security of AI systems and protecting user data will be crucial in maintaining public trust and preventing potentially catastrophic breaches.
Enhanced safety Safety is critical in transportation, and AI is emerging as a powerful tool in reducing accidents and saving lives. Advanced driver assistance systems (ADAS) and autonomous vehicle technologies use AI to detect and respond to potential hazards faster than human drivers. In public transit and freight operations, SEPTEMBER 2024
4
ITSInternational.com
A I I N T R A N S P O RTAT I O N
Ethical concerns The implementation of AI in transportation raises complex ethical questions. For instance, how should autonomous vehicles be programmed to make split-second decisions in potential accident scenarios? How do we ensure that AI-driven systems don’t perpetuate or exacerbate existing social inequalities in access to transportation? Addressing these ethical concerns will require ongoing dialogue between technologists, policymakers and the public.
networks, where self-driving vehicles seamlessly interact with smart infrastructure to optimise urban mobility. AI could enable new modes of transportation, from autonomous flying taxis to hyperloop systems, reshaping our concepts of distance and urban planning. Having said that, we must remember that technology should serve people, not the other way around. The potential benefits of AI in terms of efficiency, safety, sustainability and user experience are indeed enormous. However, these advancements must be guided by a singular focus: creating more livable cities and transportation systems that prioritise human needs and wellbeing. The future of mobility is not about optimising for machines or pursuing efficiency for its own sake. Instead, it’s about leveraging AI to create transportation ecosystems that enhance quality of life, foster community connections and respect the diverse needs of all city dwellers. This means designing AI systems that prioritise pedestrians and cyclists, enhance public spaces, reduce noise and air pollution and make our streets more vibrant and inclusive. As we navigate this transformation, it’s crucial to approach AI not as an end in itself, but as a powerful tool that, when wielded responsibly, can help us build transportation systems that truly serve people. This requires ongoing dialogue between technologists, urban planners, policymakers, and - most importantly - the communities that will be affected by these changes. The AI revolution in transportation is just beginning, and its ultimate shape will be determined not by what is technologically possible, but by what is socially desirable. As we embark on this journey, let us embrace the possibilities of AI while steadfastly ensuring that our cities remain human-centric. Our goal should be clear: to create a future of mobility that doesn’t just move people efficiently, but one that contributes to happier, healthier and more connected communities. In the end, the true measure of success for AI in transportation will not be the sophistication of our algorithms or the efficiency of our systems. It will be the smiles on the faces of children playing safely in their neighbourhoods, the ease with which elderly residents can access vital services and the vibrancy of our public spaces. By keeping these human outcomes at the forefront, we can ensure that the AI-driven transportation revolution truly serves all of humanity, creating cities that are not just smart, but also deeply livable and lovable.
Workforce impacts While AI promises significant productivity gains, it also has the potential to disrupt traditional jobs in the transportation sector. From truck drivers to traffic controllers, many roles may be transformed or eliminated by AI technologies. Managing this transition and ensuring that workers are prepared for the jobs of the future will be a critical challenge for the industry and society as a whole. Infrastructure requirements Realising the full potential of AI in transportation will require significant investments in infrastructure. This includes not only physical infrastructure like sensors and communication networks but also digital infrastructure for data storage and processing. For many cities and regions, the cost and complexity of these upgrades may be a significant barrier to adoption.
Prioritising people in the AI transportation revolution As we look to what’s in the stars for the future, the potential applications of AI in transportation seem boundless. We may see the emergence of fully autonomous transportation
AI is streamlining tolling and congestion charging systems, making them more efficient and fair
SEPTEMBER 2024
5
ITSInternational.com
AI In Transportation_8-21-24 rev2.pdf 1 8/21/2024 4:19:36 PM
SMARTER AI FOR SAFER ROADS. The Iteris ClearMobility® Platform harnesses the power of AI to revolutionize road safety, enhance travel efficiency, and support thriving communities. Here’s how:
C
Proactive Intersection Safety: Our AI-driven predictive models, powered by real-time traffic data, identify high-risk intersections and deploy timely interventions, reducing crash risks before they happen.
AI-Enhanced Traffic Management: Using advanced AI technologies, our platform continuously analyzes traffic volume and incidents, optimizing traffic flow and minimizing delays across vast networks.
Optimized Traffic Incident Management: We process billions of GPS data points daily, using AI to translate complex traffic incidents into actionable insights, improving response times and precision in traffic management.
Intelligent Video Detection: With AI-powered video classification, our systems detect and classify road users with unparalleled accuracy, ensuring safer intersections and seamless integration with connected vehicles.
M
Y
CM
Iteris ClearMobility – AI at the heart of smarter, safer travel.
MY
CY
CMY
K
Iteris.com/clearmobility
A I I N T R A N S P O RTAT I O N
How Iteris leverages AI: The Key to Smarter Mobility The company also uses this technology to provide enhanced mobility data called ClearData™ to motorists across North America, delivering real-time traffic updates via satellites, Digital Audio Broadcast (DAB), or GSM cellular networks. All this minuteby-minute real-time traffic flow information is enhanced and standardised by AI.
The integration of Artificial Intelligence (AI) in traffic management is revolutionising the way we navigate our roads and intersections. From improving real-time traffic updates to enhancing road safety and efficiency, AI is at the forefront of creating smarter, more adaptive transportation systems. Iteris is not just discussing the potential of Artificial Intelligence— it’s implementing it. Iteris’ advanced AI solutions are already enhancing the way roads are monitored, managed, and made safer through predictive modelling, real-time data analysis, advanced detection systems, and more. Read on for more details on how Iteris utilises AI and machine learning.
AI in Traffic Analytics Software AI and machine learning have serious potential to improve intelligent transportation systems (ITS) software solutions, which Iteris recognised early and made significant investments in studying and implementing. Recently, the company used AI to complete an analysis on intersection safety and came away with valuable and actionable insights that it has implemented in its technology. Iteris’ analysis of intersection safety revealed that hard acceleration predicts crashes. Intersections with high crashes also have high speeding events, and AI models outperform traditional methods in predictive accuracy. By utilising connected vehicle event data and AI models, it’s possible to proactively identify high-risk intersections and implement safety measures before crashes occur. That’s where Iteris’ software comes in: ClearGuide® uses these AI models to turn raw data into intersection rankings based on safety risks and suggested countermeasures. This heightened level of insight makes it easy for practitioners to identify high-risk areas and implement appropriate measures. This innovative approach offers timely interventions, resource optimisation, and enhanced public safety by reducing reliance on historical crash data. Iteris is proud to be at the forefront of transforming traffic management and safety with AI. Through advanced detection systems like Vantage Apex, real-time traffic management powered by ClearData, and advanced transportation software like ClearGuide, Iteris is redefining the approach to road safety and efficiency. The ongoing evolution of AI and its applications in transportation promises a safer, more efficient, and more intelligent mobility landscape—and Iteris is playing a significant role in driving these advancements.
AI in Traffic Detection Iteris has been pioneering the shift towards AI-driven smart mobility with its Vantage Apex® system, which leverages Convolutional Neural Networks (CNNs) to mirror human cognitive abilities in recognising and classifying various road users, even in challenging visual conditions. This advancement boosts intersection efficiency and safety and paves the way for seamless integration with connected and automated vehicles. But how did they get here? When Iteris began offering systems for vehicle detection using video cameras, traditional algorithms were built on machine vision technologies. Since then, Iteris’ approach has always been to continuously innovate those algorithms, ensuring detection authenticity and improving the way the systems distinguished objects. Today, traditional algorithms have reached their limits in improving detection. Recognising this, Iteris developed Vantage Apex, the industry’s first full 1080p high-definition video and 4D/HD radar hybrid sensor with integrated AI algorithms. Its AI mimics human vision using CNNs and an expansive annotation library that it has built up over the last three years. Boasting 150,000+ meticulously curated images and encompassing more than 1 million diverse objects, this collection of training data leads the field in AI at the intersection.
AI in Traffic Management Iteris transforms traffic management by utilising AI-driven telematics, processing 9 billion GPS data points daily and integrating over 150 sources. Its AI-powered ClearMobility® data engine translates complex traffic incident descriptions into numerical formats using an Expert System. This approach significantly enhances the precision and efficiency of mapping incidents to event codes.
SEPTEMBER 2024
7
ITSInternational.com
AI – A win-win situation ITS Germany’s leadership to develop AI-based environmental and mobility management that will make mobility more efficient, safe and needs-based. The technology is very promising, but knowing what AI technologies to apply and when to apply them is critical for their successful application. By leveraging our decades of traffic control experience with our talented software/AI developers, we are excited to be at the forefront of this process.
The use of Artificial Intelligence in traffic management is at an exciting phase. It enables us to address traffic flow in ever more advanced ways, with evolving detection systems, and more data accessible than ever before. SWARCO mainly uses AI support in intersection control, traffic engineering and the fleet management of public transport. “SWARCO’s public transport fleet management software NEXT integrates some machine learning algorithms to improve the quality of inputs from planning systems”, says Fabrizio Biora from SWARCO ITALIA. This refers partly to shape learning, i.e. learning by experience in order to correct and optimise routes that buses are following, and partly improving general operation by associating drivers and vehicles with shift plans. Future potentials of AI are predictive models to improve the quality of forecast at bus stops/intersections; learning models to optimise the operators’ actions to manage major service disruptions; and predictive models linked to the operation of electrically powered buses.
AI in intersection control At the core of traffic intersection, control algorithms have a series of priorities: •
•
AI in traffic engineering •
Traffic engineering and planning office SCHLOTHAUER & WAUER in Berlin, part of SWARCO, has focused on machine learning and artificial intelligence methods for several years, whether for forecasting switching times to improve connected driving (Green Light Optimal Speed Advisory) or actively influencing trafficdependent control systems to increase capacity. Their domain knowledge and increasingly sophisticated AI processes contribute to improving traffic quality, complying with pollutant limits and reducing emissions and energy consumption. SCHLOTHAUER & WAUER’s traffic experts and engineers are actively involved in AI-related research projects such as AIAMO - Artificial Intelligence And Mobility. This is funded by Germany’s Federal Ministry of Digital and Transport, in which 12 partners are working under SEPTEMBER 2024
8
Fixed time control. The signals follow the same timing repeatedly. This has been at the core of traffic signal control since its inception. Its performance hinges on the traffic engineer’s estimation of how busy the intersection will be. Vehicle actuated control. By adding detectors like loops, radar or cameras to the control, the detection input can be used to extend or cut green lights to fine-tune control. AI-based algorithms. While detection technology has advanced, traffic algorithms have not kept pace. Most detection today simplifies the data it collects in detection zones, mimicking the original loop-based detection. But detection systems and connected vehicles are capable of much more. They can deliver position, speed, classification, and more data on each vehicle or road user. This data can be used to solve the weaknesses of earlier approaches. Compared to zone-based detection, it is also a much larger quantity of data, and new approaches are needed to undertake optimal control in the sub-section signal domain. By utilising machine learning, the algorithm can measure, rather than estimate, many variables in the traffic flow, leading to much-improved control. ITSInternational.com
A I I N T R A N S P O RTAT I O N
The five advantages of SmartAI
investments to increase the performance of their signalised intersections. But it can be hard, and expensive, to check if the available products are delivering the promised improvements. Various methods can be employed to measure the effectiveness of improvements, but they often do not give the full picture. Since SmartAI can measure performance characteristics of the intersection, it can evaluate itself. By installing it first in an observation mode, where it simply measures the performance of the existing control algorithm, and then turning on the SmartAI control afterwards. road owners can see exactly what improvements they are getting.
SWARCO’s integrated approach to intersection AI control, SmartAI, gives five areas with obvious advantages compared to previous traffic control. Performance Without improved performance, there would be little call for new approaches in traffic control and extensive simulations and realworld trials are being undertaken to verify this performance. The results are impressive. In business-as-usual scenarios, where actual traffic corresponds well to the traditional traffic design, SmartAI still outperforms vehicle-actuated control by 10% or more. But the worse traffic gets, the larger the performance gains with SmartAI. We showed that increasing traffic by 20% would double delays in the intersection using traditional control algorithms. With SmartAI, the delay increased by less than 10%. In redirection events or accidents, where traffic is bottlenecked and suddenly released, SmartAI can outperform vehicle actuation by up to 80% and adaptive algorithms by 30% or more.
Complex becomes simple A tricky part of programming or configuring traditional traffic algorithms is handling special vehicles. Emergency vehicle priority and public transport priority are special modes where the traffic engineer must make decisions on how to handle these vehicles approaching the intersection. With SmartAI, anything that can be measured, can be prioritised. So, if the system can see a bus, for example, the road owner can simply place a higher value on it in the system to give it priority.
Ease of deployment AI technology can be used not only to increase performance of the intersection control, but also shorten the time spent to program, configure, and deploy it. Since SmartAI can measure traffic as it happens, it needs to know very little about the intersection in advance. It only needs to be told the safety parameters, and the priorities. It is important to note that SmartAI is not responsible for the intersection’s safety. This comes under the SWARCO ITC-3 traffic light controller. However, the AI algorithm must know under which rules it is playing in order to make the optimal decisions.
SmartAI has built-in evaluation capabilities to show exactly what the value delivered is
Insights Another aspect of the data-driven AI approach is that SmartAI can deliver high-quality traffic data to the road owner. Rather than inferring the state and performance of the intersection based on counting vehicles in zones, SmartAI measures the things road owners care about: actual delay through the intersection, number of stops and travel time. “With traditional traffic control, the road owner or traffic engineer knows pretty much how the controller will distribute the green time at the intersection,” explains Anders Palm from SWARCO TECHNOLOGY in Odense, Denmark. “With SmartAI there is a paradigm shift. Whereas traditional traffic control is all about signal timing, and the traffic being an input to this, AI-based control is all about handling the traffic, and the signals being a tool to do so. While this might seem like a minor difference, it is quite profound.” Gone are the well-known signal timing plans that can be inspected and corrected ahead of installation. The traffic control happens live, as traffic flows. It can feel uncomfortable to leave traffic flows in the hands of a machine. So, SmartAI has an inbuilt visualisation in its web interface that shows not only how the traffic is being controlled, but also gives insights into why the decisions are being made. By observing the decision-making in real time, trust is built up between road owner and controller.
Are there pitfalls? As with any new technology, AI-based intersection control removes old problems but adds its own. We are committed to help municipalities and road authorities navigate these. We have compiled a list of contingencies for road owners when considering AI-based traffic control: •
•
•
Self-evaluation Most traffic light controller installations today are replacements of existing, outdated equipment, or rebuilding existing intersections. Often municipalities and road authorities are ready to make SEPTEMBER 2024
Fallback. Since AI-based control works directly on traffic data to work, it is also more dependent on it. SmartAI uses different strategies to deal with detection outages or degradation to keep the intersection running as optimally as possible while the fault is being resolved. Evaluation is key. What is the promise being made? Does it apply to my intersection? And what was the result? It is easier to make claims than results. SmartAI has built-in evaluation capabilities to show exactly what the value delivered is. Data quality. The control system can only make decisions on available data. So, care should be taken to make sure that the detection system is optimally placed in the intersection and capable of delivering the data quality needed.
At SWARCO, the experts have no doubt: employing AI in intersection control is a win-win situation, lowering cost for road owners and minimising delays for road users.
9
ITSInternational.com
Tolling+ BORN FOR MLFF
Two AI accelerators for 99.9% of detection rate
www.tattile.com
A I I N T R A N S P O RTAT I O N
Stark OCR and Tolling+ Tattile’s artificial intelligence cameras Tattile, a leading Italian company in advanced artificial intelligence cameras, has developed a range of high-performance products and software solutions for traffic management. Its Optical Character Recognition (OCR) technology and Smart Tolling+ camera are particularly significant in areas where there is a growing need for efficient and accurate tolling and vehicle monitoring systems.
OCR algorithms Tattile’s OCR technology, embedded in ANPR cameras, ensures the precise reading of licence plates, even in challenging conditions such as high speeds or poor lighting. Tattile’s OCR systems are entirely developed in-house and deployed in over 75 countries. Thanks to the all-internal development, it can guarantee a perfect combination of hardware and software. The hardware architecture is designed to maximise the software performance and, at the same time, the software is developed to exploit the hardware architecture to its highest potential. This means that Tattile has unparalleled accuracy and a reduced execution time (up to six transits per second). Tattile’s team follows the customer from the definition of KPIs to the implementation phase. New OCR libraries can be created quickly on request, usually within three months. Tattile’s OCR technology is designed to recognise licence plates, even those that are particularly detailed, filled with tiny details or have different colours. The algorithm can also detect and recognise the presence of dangerous goods or hazardous materials by analysing ADR plates on the vehicle or its trailer. Engineered to function seamlessly in real time, it delivers immediate and accurate data by utilising advanced AI algorithms. This is especially advantageous for toll collection applications, where rapid and reliable recognition is crucial to maintaining smooth traffic flow and reducing congestion. The latest addition to Tattile’s family is the OCR developed for the US, which has a verified state and country recognition accuracy of up to 96%. Thanks to its high-end technology, driven by artificial intelligence combined with neural networks, the Stark OCR-USA has an execution time of less than 200 milliseconds and is available on every Tattile camera. The Tolling+ camera is specifically designed for multilane free-flow (MLFF) tolling applications, delivering high accuracy and performance without the need for external triggers. The camera has an impressive 99.9% detection accuracy rate and covers up to two lanes in tolling applications. Tolling+ detects transit simultaneously on the black and white OCR sensor and on the colour CTX sensor, both with the same field of view. The double detection and reading compensates SEPTEMBER 2024
for eventual exposure issues and also ensures optimal licence plate reading in harsh exposure conditions or when dealing with damaged licence plates. These artificial intelligence cameras offer a range of key features, including: •
•
•
High-resolution images: the camera is equipped with sensors capable of up to 8 megapixels on the OCR channel, covering two lanes, and providing precise and detailed images that enhance the accuracy of licence plate recognition; Dual AI accelerator: the integration of dual AI accelerators allows the camera to process large volumes of data quickly and accurately, making it ideal for high-traffic scenarios; IEC-62443 cybersecurity certification: Tattile’s cameras are all certified with IEC-62443 cybersecurity standards and receive more than 12 updates per year.
These ANPR cameras are designed to handle the complexities of modern tolling systems, adapting to diverse conditions while maintaining high performance in high-traffic areas. Their advanced AI capabilities distinguish between vehicle types, ensuring accurate toll calculation and improving the overall efficiency of toll collection systems. By providing reliable and efficient traffic management tools, Tattile is contributing to the creation of more sustainable and manageable urban environments. Tattile’s OCR technology and artificial intelligence cameras are set to make a substantial impact on traffic management and toll collection, ensuring high performance in diverse traffic management scenarios. By combining high-resolution images, advanced AI processing and reliable performance under various conditions, these products offer a comprehensive solution for modern traffic management needs. • To learn more about Tattile’s solutions, visit the company’s website at www.tattile.com
11
ITSInternational.com
A I I N T R A N S P O RTAT I O N
Edge Computing Hits the Roads: A Look at iTHEIA’s Traffic Transformation In the rapidly progressing world of transportation, the convergence of artificial intelligence (AI) and cutting-edge computational models is creating unparalleled opportunities for innovation. Edge computing, which processes data closer to the source of information, has emerged as a pivotal technology in this domain, alleviating bandwidth concerns and enabling real-time operations. This approach is especially transformative in vehicle classification and traffic management systems, where timely and accurate data is critical. Quarterhill’s iTHEIA is an example of this technological synergy, combining AI with edge computing to deliver an AIpowered vehicle classifier that provides transportation agencies with a powerful tool to collect vehicle count and classification data in a safer, more efficient way.
for temporary traffic studies and operates autonomously off a rechargeable battery. Permanent iTHEIA operates off a hardwired power source and is designed for long-term deployment.
Enhancing Safety with AI Safety is a top priority in traffic management, and the employment of iTHEIA by institutions such as Pennsylvania DoT (PennDOT) illustrates the system’s positive impact. By replacing the need to have people stationed at the roadside to count vehicles by class, iTHEIA significantly reduced the risk of putting employees in harm’s way and improved the overall safety of traffic data collection practices. Similarly, Minnesota DoT (MnDOT) adopted iTHEIA for its quick set-up and risk-free operation. According to John Hacket, supervisor of vehicle classification and traffic forecasting at MnDOT: “iTHEIA is a game-changer for us. Deploying this AI classifier device is fast and simple—set-up in under 30 minutes— and it watches over three lanes of traffic without risk to our staff’s safety on busy roads. The iTHEIA AI Classifier is the only way we can get FHWA 13 Class counts on some of our busy roadways.”
Transformative Technology for Traffic Management Quarterhill has positioned iTHEIA as a quintessential tool in the traffic management arsenal, with its ability to harness edge computing for efficient local data processing. Unlike some videobased traffic data systems, all vehicle classification is done at the roadside without requiring video to be uploaded to the cloud for processing. This means that iTHEIA can be deployed in areas with limited or no internet connectivity, and also reduces the amount of data that needs to be transmitted over the network. This makes it a more cost-effective solution for transportation agencies looking to collect traffic data at remote locations, or for short periods of time. It does, however, offer an API and the option of connecting it to other systems, so it can be used as part of a larger traffic management solution.
A New Era of Traffic Data Collection iTHEIA’s implementation marks a step forward in traffic data collection, one that no longer relies on in-road sensors or other intrusive methods and instead transforms video into a precise, efficient vehicle classification method right at the site. This advancement is emblematic of a broader move towards smarter, data-driven infrastructure management decisions.
Striving for Precision and Efficiency iTHEIA’s data meets or exceeds Transportation Management Guide standards for count and class detection. Using the Federal Highway Administration (FHWA) 13-Class scheme for traffic categorisation, the system gathers comprehensive data on vehicle volume, type, lane usage and speed. Gathering such information is vital for the development of informed traffic management strategies and the planning of infrastructure enhancements. iTHEIA is capable of monitoring up to four lanes with a single camera, or eight lanes by using two, demonstrating remarkable efficiency. Installations can often utilise existing infrastructure, like roadside poles, minimising disruptions and reducing installation costs. Two versions of iTHEIA are available for adaptability across various project requirements. Portable iTHEIA offers versatility SEPTEMBER 2024
Customisation and Global Adaptation iTHEIA can be tailored to fit various national traffic classification standards, providing a flexible and indispensable tool for agencies internationally seeking precision and adaptability in their traffic management frameworks. In summary, edge computing has proven its potential in transportation, particularly through AI-driven platforms like iTHEIA. This state-of-the-art system showcases what can be achieved when AI and edge computing converge, propelling the industry from aspirational vision to tangible reality.
13
ITSInternational.com
Powerful Cloud-based ATMS with Predictive Traffic Forecasts using Artificial Intelligence Delivering New Levels of Traffic Signal Control and Intelligent Automation
Econolite’s Centracs® Mobility Platform, the industry-leading, secure, flexible, cloud-based Advanced Transportation Management System (ATMS) now integrates PTV Flows, the predictive
traffic management solution that leverages Artificial Intelligence (AI) to forecast travel times, enabling proactive signal control decisions. Centracs Mobility, now with PTV Flows, is the ONLY ATMS capable of moving traffic signal optimizations from reactive to proactive. Our new Traffic Proactive optimization module is layered on top of existing capabilities for data analytics and reporting, signal/detection management, additional optimization services, and much more.
umovity.com
Powered by
econolite.com
ptvgroup.com
A I I N T R A N S P O RTAT I O N
Artificial Intelligence: The Breakthrough Needed the quality and speed of decisions made in the face of dynamic traffic conditions. It will also establish a platform where predictive analytics and forecasting can support future planning and optimisations.
It’s no secret that reducing traffic congestion and enhancing safety on our roadways present an increasingly complex challenge, with widespread implications. The negative economic and environmental impacts due to congested roadways have been well-documented, from loss of productivity and jeopardised commerce to increased fuel consumption, vehicle emissions, and risks to vulnerable communities. Adding to the complexity is that roadway safety continues to be a persistent challenge and is often addressed independently of traffic congestion. Recent innovations have helped to incrementally address roadway safety. But even with today’s state-of-the-art ITS solutions in place, these systems are designed to adapt or react to roadway conditions. What is really needed to make substantive and sustainable gains is a breakthrough to forecast and predict problems before they occur, so that appropriate real-time adjustments can be made to prevent those problems from happening. Fortunately, ongoing developments in critical hardware and software technologies are delivering key situational awareness capabilities and transforming traffic and transit management control. The convergence of technologies is finally making Artificial Intelligence (AI) a reality and delivering predictive analytics and advanced machine learning as practical tools for optimising traffic operations. We now have the necessary resolution of historical and realtime data to fuel new algorithms within AI to forecast and predict traffic conditions. AI-driven applications can continually help optimise roadways to prevent congestion and identify emerging safety risks associated with traffic conditions, weather and user behaviours. These capabilities will have profound positive impacts on the future of traffic operations and delivery of safe and sustainable mobility.
Bridging the Vision Zero Gap Leveraging AI can potentially support very significant advancements toward Vision Zero. AI-driven safety enhancement applications can combine the latest sensing and situational awareness technologies with the elements of prediction and proactive action to introduce new levels of efficiency in the forecasting, detection and mitigation of safety-related incidents, as well as incident response. This is especially true for the safety of Vulnerable Road Users (VRU) like pedestrians and bicyclists. AI-powered safety applications can help save lives and mitigate safety risks for both roadway users and first responders. The use of AI, combined with powerful analytics based on historical and real-time traffic, weather and incident data, enables early identification and efficient mitigation of safety risks. Continuous safety monitoring along with micro and macro simulation also enables infrastructure owners and operators to determine and implement appropriate pre-emptive planning measures to improve safety with an eye on the goal of Vision Zero.
What’s Next? Applying AI tools to traffic management will depend on the commitment and efficacy of municipalities towards adopting and deploying advanced ITS technology as part of their nextgeneration transportation infrastructure. Applied properly, AI holds the promise of being a game-changer in holistically optimising traffic operations and delivering more sustainable and safer mobility.
AI Enhancing Traffic Management Operations In addition to providing smarter machine vision and improved situational awareness, it is key for AI-driven systems to incorporate both historic and real-time traffic data to forecast traffic conditions before they happen and combine this with a powerful Advanced Traffic Management System (ATMS) to apply proactive traffic signal control in real time. This dynamic and forecasting-based approach to traffic control helps mitigate traffic bottlenecks before they occur. This moves traffic signal optimisation from reactive to proactive by leveraging real-time situational awareness, travel-time predictions, and AI to dynamically optimise traffic and transit management operations. Another important aspect of leveraging AI will be its use in Digital Twin and advanced Extended Reality (XR) technologies. Enhancing transportation management solutions powered by AI in a Digital Twin environment will provide forecasting, simulation, and validation capabilities, improving SEPTEMBER 2024
15
ITSInternational.com
From the Past to the Future: Building the Roadway to Artificial Intelligence fundamentally flawed, potentially leading to costly mistakes and inefficiencies. Trust in data isn’t just desirable - it’s essential for the successful implementation of AI in transportation management.
Our transportation systems are in crisis. Every 10 minutes, $100 million vanishes - lost to traffic fatalities, wasted time, and pollution. That’s $5 trillion annually. Despite the US government pouring more than $350 billion yearly into transportation infrastructure, our roads remain congested and our air polluted. Meanwhile, traffic engineers are being left to fight an impossible battle, sent to a gunfight with nothing but a pencil and a clipboard. AI-powered tools will level the playing field.
The Evolution of Transportation Data Data sources have been growing over the last several decades – including sensors in roads, cameras on traffic lights, GPS data from vehicles, and smartphones. They all generate constant streams of information. Yet, paradoxically, much of the work and collection of the data in transportation management is still done manually or with outdated methods. The biggest problem with all this data, however, is ensuring that it’s accurate, comprehensive, and properly utilised.
The Promise of AI Artificial Intelligence (AI) holds immense promise for revolutionising transportation. We’ve seen its impact in consumer applications like ride-sharing apps and autonomous vehicles. However, when it comes to the fundamental management of our transportation infrastructure, AI’s potential remains largely untapped.
The Data Dilemma: Quantity, Quality and Integration
Every Decision Starts with Data
Bigger Data is Better Data Despite the recent growth in collection methods, there simply isn’t enough accurate data that agencies can access to get a comprehensive view of their transportation networks. Sensor technologies are often too expensive to deploy widely, leaving significant blind spots in our understanding of traffic patterns and road conditions. Probe data and connected vehicle data have been used for over a decade, but with penetration rates averaging less than 5%, this data is often too sparse to provide meaningful insights, especially on local and rural roads, or during low-traffic periods.
To understand why AI hasn’t yet transformed transportation management as dramatically as other sectors, we need to grasp a fundamental concept: every decision starts with data you can trust. This axiom takes on critical importance in the age of intelligent systems. It’s not just about having data - it’s about having accurate, reliable data that engineers can stake their decisions on. Historically, transportation engineers have had to go out into the field to directly observe traffic patterns. It was the only way to get data they could truly trust. This method, while time-consuming and limited in scope, provided first-hand, verifiable information that engineers could confidently use in decision-making. The quality and trustworthiness of data directly impact the quality of decisions made based on it. In AI, this principle is amplified exponentially. If the data is inaccurate, unreliable, or untrustworthy, the resulting decisions and predictions will be SEPTEMBER 2024
Better Data is Better Data Even where abundant data exists, reliability is a significant issue. In our detection quality studies, we’ve found that less than 27.6% of detection devices are providing reliable data and more than
16
ITSInternational.com
A I I N T R A N S P O RTAT I O N
Our AI uncovers insights that surpass human observation capabilities. We generate high-definition maps with centimetrelevel precision, detect asset wear through video analysis and assess road conditions using connected vehicle data. By leveraging smartphone information, we gain crucial insights into pedestrian and cyclist movements. This multifaceted approach reveals complex traffic patterns invisible to single data sources, allowing us to understand roadway dynamics in unprecedented detail. By creating a digital twin of the transportation network, we’re offering agencies unparalleled accuracy and insight because it’s not just about seeing the road—it’s about comprehending the intricate dynamics of our transportation systems.
50% of signal operations issues are a result of malfunctioning or unreliable detection. Improving the quality and accuracy of data is paramount. Without a solid foundation of accurate, reliable data, even the most sophisticated AI models will produce flawed results. Integrated Data is Better Data While having more high-quality data is crucial, it’s equally important to recognise that too much of the same data can be problematic. Every dataset has its strengths and weaknesses, and no single data source can provide a complete understanding of traffic conditions. In the transportation ecosystem, we have multiple sensor providers, numerous systems managing roadways, and hundreds of millions of vehicles. Each of these sources provides valuable information but, too often, this data is stuck in silos, owned by different entities and stored in proprietary formats.
Building the Bridge to the Future These foundational steps have enabled us to build the bridge to the future of transportation. By ensuring that both engineering teams and AI systems can access this high-quality, integrated data, we’ve created a solid foundation for advanced AI applications in transportation management. With this foundation in place, we’ve already developed and deployed several sophisticated AI solutions:
Building the Future of Transportation At Flow Labs, we’ve been focused on addressing these fundamental challenges of data quantity, quality, and integration. Unlocking Data Silos We’ve revolutionised our data collection approach by partnering with OEMs, GPS manufacturers, and insurance companies. This collaboration allows us to capture real-time data directly from vehicles, achieving an unprecedented penetration rate of over 43% of all vehicles in the US, a significant increase compared to the current industry standard of 3-10%. To complement this, we’ve partnered with ITS device manufacturers to build custom interfaces with various ITS devices, opening up vast volumes of data that were previously trapped. The combination of this extensive vehicle data and newly accessible ITS device information provides a more complete and dynamic picture of our transportation networks than ever before.
•
•
•
AI-Powered Data Integration and Cleaning Our AI technology is revolutionising how we understand and manage our roadways. We integrate diverse datasets into a single, comprehensive geospatial data structure, using advanced AI algorithms to cross-reference and validate information from multiple sources. This process allows us to quickly identify and eliminate inaccuracies, resulting in datasets with over 95% accuracy for any given roadway.
•
Traffic Signal Optimisation: We’ve transformed a process that typically takes engineers up to 70 hours per intersection into one that can be completed in seconds. The results are impressive: a 24% reduction in travel times, 21% reduction in emissions, and 51% reduction in crash risk. AI Digital Twins: We’ve created detailed, accurate simulations of roadways, allowing engineers to predict and understand the impact of their decisions before implementation. Safety and Congestion Management: Our AI models are already assisting in roadway safety prioritisation, congestion prediction and diversion planning, helping agencies proactively address issues before they become critical problems. Predictive Maintenance: By analysing patterns in our integrated data, we’re now able to predict when infrastructure is likely to fail or require maintenance, allowing for more efficient resource allocation and reducing unexpected disruptions.
These applications represent significant advancements in transportation management, but they’re just the beginning. As we refine our processes and AI models, we’re paving the way for even more sophisticated solutions. Advanced applications like real-time adaptive systems for dynamic traffic management, multimodal transportation planning tools, and emissions reduction strategies are not just pipedreams, they’re part of the roadmap. We’re excited to tackle these complex transportation challenges and push the boundaries of what’s possible with AI in urban mobility.
The Future of AI in Transportation The potential for AI in transportation is not just vast—it’s transformative. Today, we stand at the threshold of a new era in urban mobility, where AI can help us create transportation networks that are cleaner, clearer, and safer – for everyone. Are you ready to be part of this exciting journey towards an AIpowered transportation revolution? www.flowlabs.ai SEPTEMBER 2024
17
itsinternational.com
RELIABLE DETECTION RADAR + CAMERA + A.I. = MORE RELIABLE LOWER COST Proven Performance Over 50,000 radar units installed Cost-Efficient Lower cost than camera detection Comprehensive Detection Stop bar and Advance detection Pedestrian, bicycle, and vehicle classification Compact Design Smallest cabinet footprint Advanced Capabilities ATSPM and NTCIP reports
Addressing Your Biggest Challenges SOLVE SIGNAL MALFUNCTIONS Stop dealing with unreliable systems — our technology ensures improved detection accuracy and lower false detection rates ENHANCED DATA ANALYTICS AND REAL-TIME MONITORING Get the data you need with our advanced reporting features and real-time monitoring, allowing you to make data-driven decisions to optimize traffic flow USER-FRIENDLY INSTALLATION AND SUPPORT Worried about training and installation errors? Our system is designed to be simple to install, configure and integrate into your current setup and provides configuration verification and diagnostics to ensure just one short site visit
Contact us: www.radarvision.ai
A I I N T R A N S P O RTAT I O N
Radar Vision’s Deep Learning Solves Traffic Congestion Engineers at Radar Vision have solved traffic congestion through Deep Learning of trajectory data produced by low-cost digital beam steering radar and analysis of human behaviour.
subtle cues, embedded in the trajectory data, offer critical insights into how drivers perceive their traffic experience. By applying Deep Learning algorithms to this data, Radar Vision has developed a model that can predict human experience and optimise traffic signals based on how drivers are likely to feel. The large-scale, precise data captured by digital beam steering radar provide the necessary dimensionality and precision to train these models effectively. The result is a system that optimises signal timing based not just on raw traffic numbers but also on the psychological and behavioural responses of drivers.
The Problem with Traditional Traffic Detection Methods Traditional signal optimisation has relied on detection technologies like inductive loops and video cameras to monitor presence at intersections. These systems have several limitations, notably poor detection accuracy and inability to measure key parameters like long queue length, queue growth rate, lane change percentage, speed, acceleration, response to signal change, dwell time and gap. The data from such systems are often sparse and unreliable, especially in variable traffic conditions, leading to suboptimal signal timing and increased congestion. Advanced video detection also suffers from short range compared to radar. Additionally, public dissatisfaction with traffic signals is a growing issue. Motorists often feel that they are unfairly treated when compared to other vehicles at intersections, particularly when they experience longer waits than other lanes or other approaches. This sense of congestion is not just about how long drivers spend on the road but about their perceived experience. The inability of traditional systems to measure these human factors leaves traffic engineers with a service delivery problem, hampering efforts to improve signal coordination and optimise traffic flow using new technology.
Optimising Traffic with Evolutionary Computing Since feelings cannot be accurately surveyed at traffic signals, Deep Learning makes this system ideally suited for unsupervised learning and evolutionary computing. Unsupervised learning allows the model to continuously improve without human intervention, adapting to new traffic patterns and behaviours as they emerge. Evolutionary computing, a type of optimisation algorithm that mimics the process of natural selection, ensures that the system is constantly evolving to provide the best possible signal coordination and to maximise public perception of service delivery. This dynamic approach allows traffic systems to respond to real-time conditions with unprecedented accuracy, reducing congestion, improving traffic flow, and enhancing public satisfaction. By understanding the human element of traffic management, Radar Vision’s solution creates an evolutionary system that optimises signals based on both objective data and measured human satisfaction.
Introducing Digital Beam Steering Long Range Radar
Conclusion
Engineers at Radar Vision have developed a cutting-edge solution to address these challenges: Digital Beam Steering Long Range Radar, coupled with Deep Learning algorithms. This advanced radar technology offers precise behavioural characteristics such as gap acceptance, dwell time, deceleration, lane adherence and queue joining speed across a large area. These trajectories monitored by beam steering radar provide a rich dataset of behavioural information that is critical for improving traffic signal coordination through analysis and prediction, but most importantly it provides the dataset required for AI to calculate the best solution. The key to this technology’s success is its ability to accurately capture data metrics containing the human factors that affect satisfaction.
Digital Beam Steering Long Range Radar, combined with Deep Learning, provides a powerful, innovative solution that goes beyond mere vehicle detection. By analysing the behaviour and psychology of drivers, this system offers a more holistic approach to traffic management, one that can reduce congestion, improve safety and ultimately improve public satisfaction.
The Power of Deep Learning and Behavioural Analysis Traffic congestion is often defined by motorists as an unnecessary delay or a sense of unfairness at intersections. This subjective feeling can be difficult to quantify, but the trajectory data captured by long-range radar provides a wealth of information that can be analysed to understand and predict driver behaviour. For example, impatient drivers may exhibit more aggressive behaviour, such as faster acceleration or harsher braking. These SEPTEMBER 2024
19
ITSInternational.com