Adaptive Signal Coordination for Heterogeneous Traffic Using Downstream Detection

Real time control of heterogeneous traffic characterised by mixed vehicle type and non-lane based movement is always a challenge for efficient and effective traffic management. Adaptive traffic signal controllers offer better signal time management especially when the traffic is fluctuating. Traditional pro-active systems such as UTOPIA, PRODYN, SCOOT, OPAC, RHODES, etc. uses prediction models using upstream vehicle detection and works well in lane-based traffic. However, this approach has several limitations under heterogeneous traffic conditions due to inaccurate vehicle detection and prediction models. Therefore, adaptive control systems may work better with stop-line detectors and models that do not require predictions. When such systems are used in a corridor, it is important to synchronise adjacent signals to enhance the corridor performance. The synchronization is possible only by operating the junctions at a common cycle. Therefore, a traffic adaptive control system using stop-line detector information is proposed in this paper. The proposed model aims at real-time estimation of common cycle time through actor-critic reinforcement learning; an approach originated from the machine learning community. This approach has the ability to learn relationships between control action such as cycle time and their effect on the vehicle queuing while pursuing the goal of maximizing intersection throughput. In order to evaluate the system, the proposed algorithm is interfaced with a traffic simulator (VISSIM) which supplies the detector information and uses the cycle time provided by the model. The model performance is compared with the conventional vehicle actuated system. The results using this approach shows significant improvement over vehicle actuated control, especially along the coordinated direction.