In urban transportation, using artificial intelligence (AI) in traffic management is increasingly becoming crucial to reduce travel times and vehicular emissions by eliminating the bottlenecks that create traffic snarls in cities. This article discusses the importance of AI in traffic management, the applications of the technology, and recent developments.
Image credit: metamorworks/Shutterstock
AI’s Importance in Traffic Management
The primary objective of urban transport systems is to ensure reliable, efficient, and safe transportation while minimizing the environmental impact. However, traffic congestion has become a major problem over time in most cities due to the ever-rising number of vehicles on the world’s roads without a corresponding increase in road capacity.
The problem has led to increased air pollution, noise pollution, travel times, and significantly higher fuel consumption. Drivers are also experiencing chronic stress and physical health issues as they spend more time on roads due to congestion. Additionally, traffic congestion has increased accidents, such as rear-end collisions and deaths, due to higher speed variance among vehicles between and within lanes and erratic driving behavior of drivers.
Although cities are heavily investing in upgrading their existing road transportation infrastructure to reduce congestion and ensure efficient traffic management, these steps have not effectively addressed the congestion issues, necessitating the adoption of AI-based intelligent and modern systems.
The rapid advancements in AI technology have substantially increased its importance in road traffic management. AI can accurately control and predict the flow of vehicles and people at multiple points on the transportation network. Moreover, AI can decrease accidents by optimizing vehicular flow across intersections and during road closures due to activities such as construction. AI can also analyze and process substantial amounts of data to implement effective mass transit, such as ride-sharing services.
Traffic law enforcement, dynamic traffic light sequencing, intelligent parking planning, automatic distance recognition, improved urban planning, and quick passage to emergency vehicles are the major advantages of AI in traffic management. However, the use of AI in traffic management has several challenges, including a proper understanding of the underlying issues, data acquisition, feature extraction, and data processing for predictive modeling, model deployment/monitoring/updating, learning from past mistakes, and feedback analysis, integrating different data types such as image and video, scalability, standardization, cost-effectiveness, and privacy concerns.
Major AI-based Traffic Management Solutions
Intelligent Traffic Management System (IMTS): Traffic rule-breaking and rash driving significantly increase traffic congestion due to the disruption of normal traffic and also lead to accidents.
IMTS can be used to collect real-time data from traffic cameras and analyze the data using AI algorithms to detect and fine vehicles violating traffic rules, such as breaking the speed limits.
The fine can be sent electronically to the offender using the system, which reduces the need for manpower, paperwork, postage, and traffic wardens, leading to greater efficiency in traffic law enforcement.
AI Monitoring System (AIMS): AIMS can monitor the vehicle type, the number of on-road vehicles at a particular moment, vehicle speed, and the amount of traffic at any intersection using a high-definition closed-circuit television (CCTV) camera.
Thus, the system can predict traffic jams, divert vehicles to less-congested roads, reduce congestion by extending the green light duration, and provide over-speeding tickets.
Artificial Neural Network (ANN)-based Intelligent Traffic Congestion Forecasting System (ITCFS): These ANN-based systems can be used to determine the short-term and long-term traffic condition/congestion on the road and guide drivers to alternative routes to avoid congestion and ensure an efficient traffic flow.
Intelligent Sensor-based Traffic Signal Management System (ITSMS): This form of AI can be utilized to ensure the safety of pedestrians while they are crossing roads. The intelligent sensor in this system detects the number of individuals intending to cross the road when the pedestrians press a button on the traffic signal and adjusts the red light duration for the vehicles. Thus, ITSMS can significantly reduce congestion due to illegal crossings and fatal accidents.
AI-based Approaches for Urban Road Traffic Management
A multi-agent reinforcement learning-based urban traffic route guidance method with high adaptive learning capability in various traffic scenarios can be used to minimize road congestion. The method consists of two algorithms: modular Q-learning (MQ) and trajectory A*Rejection.
MQ algorithm accurately estimates the road trip cost based on the temporal and special characteristics of several traffic scenarios, while trajectory A*Rejection determines the effective path for every vehicle using the road network congestion index to ensure efficient traffic flow.
Several studies have proposed different AI approaches to minimize congestion at intersections that are based on traffic data collection through wireless sensor networks (WSN), cameras and image processing, and vehicular ad hoc networks (VANET). For instance, the supporting infrastructure of the distribuTed and AdaPtive IntersectiOns Control Algorithm (TAPIOCA) that can address the problem of controlling traffic lights at an intersection is a hierarchical distributed WSN architecture,
The architecture contains a first layer of sensors, designated as Before Light, which continuously collects the vehicle arrivals at the intersection, and a second layer of sensors, designated as After Light, which collects the vehicle departures from the intersection and only works when the traffic light is green.
The After Light sensors also aggregate the data collected for every movement, execute the decision-making process, and transmit the traffic light phase calculation results to different network sensors and the traffic light controller (TLC). Vehicle-to-vehicle (V2V)-based virtual traffic light (V3TL), a dynamic distributed solution, can create a cyclical planning process for vehicles approaching the intersection to reduce the waiting time. The solution minimizes the number of actions needed to make the intersection congestion-free and the number of go and stops of every vehicle to decrease emissions.
Intelligent traffic light controlling (ITLC) algorithm can be utilized for intelligent traffic light phase planning of an isolated intersection. This algorithm depends on the real-time traffic flow characteristics, such as the number and speed of vehicles and direction of travel, and on the results of the Efficient COngestion DEtection (ECODE) protocol, which detects traffic congestion.
Similarly, arterial traffic light control, an ITLC algorithm adaptation for open network scenarios, is based on the scheduling reports of neighboring traffic lights that are located on the arterial street. These reports are primarily delivered using vehicles traveling along the arterial traffic flows.
A fuzzy logic-based approach can be employed to effectively control traffic at intersections. In this approach, a smart camera is used for every intersection lane to monitor and assess traffic properly in real-time and track and detect special vehicles.
Recent Studies
The increasing infrastructure needs and the emergence of big data streams generated by surveillance feeds, smart sensors, and the Internet of Things (IoT) have substantially transformed the technological landscape of intelligent transport systems (ITS). Thus, ITS must harness the potential of AI to develop smart traffic management solutions driven by big data for effective decision-making.
However, the existing AI methods that function in isolation cannot facilitate the development of a comprehensive platform owing to the big data stream dynamicity, high-frequency generation of unlabeled data from heterogeneous data sources, and traffic condition volatility.
In a paper recently published in the journal IEEE Transactions on Intelligent Transportation Systems, researchers proposed a smart traffic management platform (STMP), an expansive, intelligent traffic data integration and analysis platform based on unsupervised online incremental machine learning, deep reinforcement learning, and deep learning to address these limitations.
The STMP platform contained three functional layers, designated as L1, L2, and L3. The heterogeneous data sources were transformed and integrated into the L1 data transformation layer. In the L2 layer, the online incremental machine learning algorithm detected the concept drifts and distinguished them into non-recurrent and recurrent traffic events, which were then passed as input to smart traffic management modules in the L3 layer for impact propagation estimation, traffic flow forecasting, commuter emotion analysis, and optimization for intelligent traffic control.
The proposed platform can capture the dynamic patterns from traffic data streams and integrate the AI modules for real-time adaptive traffic control and traffic analysis. Researchers assessed the effectiveness and feasibility of the platform by performing several experiments on real-time social media data and Bluetooth sensor network data from the arterial road network in Victoria, Australia.
The experimental results demonstrated that the platform could effectively detect recurrent and non-recurrent traffic incidents successfully in a timely manner. These results were also validated using the insights captured automatically from social media. Additionally, the traffic flow prediction and impact propagation modules efficiently predicted the short-term impacts of the traffic events.
Reference and Further Reading
D. Nallaperuma., Nawaratne, R., Bandaragoda, T., Adikari, A., Nguyen, A., Kempitiya, T., Silva, D. D., Alahakoon, D., Pothuera, D. (2019). Online Incremental Machine Learning Platform for Big Data-Driven Smart Traffic Management. IEEE Transactions on Intelligent Transportation Systems, 20, 12, 4679-4690. https://doi.org/10.1109/TITS.2019.2924883
Ouallane, A. A., Bahnasse, A., Bakali, A., Talea, M. (2021). Overview of Road Traffic Management Solutions based on IoT and AI. Procedia Computer Science, 198, 518-523. https://doi.org/10.1016/j.procs.2021.12.279
Top Five AI-Based Smart Traffic Management Solutions [Online] Available at https://mkai.org/top-five-ai-based-smart-traffic-management-solutions/ (Accessed on 25 August 2023)
A Comprehensive Guide on AI in Traffic Management [Online] Available at https://www.matellio.com/blog/ai-traffic-management/ (Accessed on 25 August 2023)