A multi-objective agent-based approach for road traffic controls: application for adaptive traffic signal systems

Agent-based approaches have gained popularity in engineering applications, but its potential for advanced traffic controls has not been sufficiently explored. This paper presents a multi-agent framework that models traffic control instruments and their interactions with road traffic. A multi-objective Markov decision process is applied to model agent operations, allowing agents to form a decision in the context of multiple policy goals. The problem is reformulated by a constrained Markov decision process (CMDP) to enhance the computational efficiency. In the study, the policy goal with the highest priority becomes the optimization objective, but the other objectives are transferred as constraints for optimization. A reinforcement learning based approach is developed with different function approximation methods used to enhance the control algorithm. For implementation of multi-objective control, a threshold lexicographic ordering method is introduced and integrated with the learning algorithm. While the multi-objective intelligent control method could be potentially applied for different road traffic controls, this paper demonstrates a case study on traffic signal control in a road network in Stockholm. Intersections are modeled as agents that can make intelligent timing decisions according to the detected traffic states and update their knowledge from system feedback. The evaluation results show the benefits offered by the control approach especially when multiple policy requirements are introduced.QC 20180514