In this paper, we study the optimization of traffic light controllers and present an adaptive, predictive, and statistical optimization algorithm that performs dynamic queue length estimation. The system presented focuses on low power consumption, easy maintenance, and simple construction. The highlights of the system are (1) dynamic queue length estimation for timer delay computation and (2) the signal coordination algorithm it employs. Adaptive logic focuses on estimating the queue length during run time using sensors. The sensors need not be activated if a pattern is observed in the traffic flow. This forms the substratum for the predictive logic. Statistical data is used when the queue length exceeds a threshold. The green time for each traffic signal can be varied between a pre-estimated minimum and maximum, depending on the traffic flow. The red time for a particular signal depends on the green time of its complementary signal. The queue length detectors that we propose to use are fundamentally sensor networks that are composed of through-beam photoelectric sensors, arranged in an efficient topology. The efficiency of the algorithm has been estimated by conceptually applying the algorithm to a busy intersection in Chennai, India. The related statistical comparison with current systems has been presented. The algorithms have been simulated using a computer program written for the Turbo C++ compiler. An optimized signal coordination algorithm is presented that utilizes an online timing update technique for efficient traffic flow.
Abstract