engleski

Development of Adaptive Traffic Signal Control Systems for Urban Environments

With the increase of population in urban environments, safety issues and congestion in road traffic occurred. The first traffic light was introduced in 1868 in front of the British parliament in London to solve the safety problem. It was invented by the railway engineer J. P. Knight, it was humanly operated and used two colored plates to control the traffic. With the first three-colored traffic light and corresponding controller, a new era of Adaptive Traffic Signal Control (ATSC) started with the ability to solve congestion-related problems also. A constant change of signal programs according to the current traffic situation was potentially possible from that moment. While the control devices from the first generation (1GS) of ATSC systems could only change between several predefined signal programs in predefined periods during the day (morning and afternoon peak period, selected non-peak periods), the second generation (2GS) dynamically adjusted the parameters of the signal timing scheme (signal period, green signal ratio, and phase difference) in a centralized control architecture. The third generation (3GS) added a distributed control architecture that required very limited fine-tuning to achieve optimal behavior. Further development resulted in the fourth generation (4GS) featured with an integrated traffic management and control system containing dynamic process models of combined traffic assignment and control with different signal updating strategies. Nowadays, the fifth generation (5GS) is being used with the main characteristic that the ATSC system learns the traffic control knowledge independently and reduces the computational burden of decision optimization intelligently. With the application of cloud computing, all available traffic state-related measurements (inductive loops, cameras, GNSS trajectories of buses, taxis, and emergency vehicles) and information (public transport, emergency services, social media) are integrated and signal programs optimized on the global level of an urban area. The optimization of signal programs also includes public transportation and emergency vehicles priority, forecast of the occurrence of possible traffic problems, detection and prevention of traffic rules violation, and informing about optimal routes resulting in significant improvement of the throughput of the managed road network. With the application of artificial intelligence, i.e., learning capabilities, the question now arises, will the traffic controller perform well also in situations it has not encountered in its learning process? Because methods based on machine learning (fuzzy logic rules optimized using the genetic algorithm, (self-organizing) neural networks, (deep) reinforcement learning) can guaranty optimal control behavior only for traffic situations successfully solved during the learning process. Human supervision and intervention in new unforeseen circumstances to guide the learning process is still and will be necessary.

Intelligent Transportation Systems, Adaptive Traffic Signal Control, Artificial Intelligence, Urban environments

nije evidentirano