A Feasibility Study on Upgrading the Static TLC Infrastructure to Adaptive TLC

This paper evaluates the feasibility of upgrading the static traffic light control to a local adaptive traffic light control, for a road network carrying less-lane-disciplined, heterogeneous traffic. We analyze the performance of a few deep learning based object detection algorithms (e.g., SSD, RCNN), with respect to the computation requirements, and accuracy for computing the Passenger Car Unit (PCU) count, under heterogeneous traffic condition. We propose an algorithm for a local adaptive TLC, leveraging the existing infrastructure. This algorithm efficiently computes the phase duration, based on round-robin scheduling, considering real-time traffic information. Simulations are carried out to analyze the effect of varying error rates in PCU count on the performance of adaptive TLCs. Further, the performance of the proposed TLC is compared with the conventional static TLC and the recently proposed micro auction based adaptive TLC algorithms. The simulation results suggest that the proposed TLC algorithm can tolerate 20% error in the PCU count without degrading the performance. Also, this work demonstrates that the traffic information with the required accuracy can be processed in real time using the available platforms (e.g., Raspberry Pi). The proposed work establishes the feasibility of upgrading the existing static TLC to a local adaptive TLC with minimal infrastructure requirement.