Control experiments for a network of signalized intersections using the ‘.Q' simulator

Abstract

The control of a network of signalized intersections is studied using a discrete-event simulator called ‘point queue' (.Q). Vehicles arrive at entry links from outside the network in a continuous Poisson stream, independently make turns at intersections, and eventually leave from exit links. There is a separate queue at each intersection for each turn movement. The control at each intersection determines the amount of time that each queue is served within each cycle. A vehicle arriving at an intersection joins the appropriate queue, waits there until it is served (its ‘green light' is operated), then travels over the downstream link and joins the next queue or leaves if it is an exit link. The performance of the control scheme is measured in terms of the length of each queue, the queue waiting time, and the travel time from entry to exit. Two sets of control policies are modeled and compared via .Q simulations for a fairly complex arterial network near the I-15 freeway in San Diego, CA. The first is ‘fixed time (FT)' control which generates an open loop periodic sequence of green light operations. The second is a feedback strategy called ‘max pressure (MP)' in which the turn movement that is operated is a function of the queue lengths adjacent to the intersection. The simulations confirm the theoretical property of MP, namely that it maximizes throughput, whereas FT does not. The simulation study provides more details concerning the queue length distribution and the behavior of MP as a function of how frequently it is invoked. These details are critical in evaluating the practicality of MP. The study shows that the .Q simulator is a versatile tool in the design of signal control.