OPTIMIZING REAL-TIME TRANSIT PRIORITY IN COORDINATED TRAFFIC NETWORKS WITH GENETIC ALGORITHMS

摘要

Significant difficulty in providing preferential treatments to transit vehicles along urban arterials is that buses often travel outside general traffic progression patterns (due to boarding and alighting passengers) thus creating conflicts between green time needs of general traffic and those of transit traffic. Simultaneous priority requests also complicate the green time allocation decisions, and so does the variability in boarding and alighting times. Genetic Algorithms (GA) and Artificial Neural Networks (ANN) are combined in this research to implement efficient transit priority within real-time coordinated network. An ANN model first-predicts the amount of time lost to dwelling activity at bus stops. A GA then optimizes signal timings. Microscopic traffic simulation is used to evaluate the proposed procedure. The results show the ability of the proposed model to improve traffic network performance within a coordinated system compared to the current practices where transit signal priority (TSP) is implemented within fixed-time signal systems using the mean value of boarding/alighting times.