A Methodology for Obtaining Signal Coordination within A Distributed Real- Time Network Signal Control System with Transit Priority

Real-time network signal control offer the potential to provide delay benefits over traditional fixed time and actuated control. The SPPORT (Signal Priority Procedure for Optimization in Real-Time) model is a fully distributed heuristic rule-based signal control method that explicitly considers transit priority. While a distributed architecture enables the network control problem to be decomposed in such way that local controllers can optimize individual intersections, it also prevents the explicit development of coordination along signalized corridors. In this case, coordination can only be achieved when local controllers are instructed to consider the timing plans of adjacent controllers, the vehicle departures from upstream adjacent intersections, and the projected vehicle arrivals at downstream adjacent intersections. This paper describes a coordination methodology that was developed for use within the SPPORT model to allow it consider traffic progression objectives. In this methodology, specific considerations are given to coordination with downstream signal timings, downstream queues, downstream transit activities, upstream signal timings, upstream queue spillback events, and upstream transit activities. An evaluation of the resulting model for a five-intersection arterial corridor with scenarios considering a range of traffic conditions finally shows the benefits that the application of the model can, particularly over fixed-time control. F. Dion and B. Hellinga 2 INTRODUCTION Over the past decades, numerous efforts have been devoted towards the development of efficient traffic responsive signal control methods. Examples of such methods are the OPAC (1-3), Rhodes (4-6), SCATS (7), SCOOT (8), PRODYN (9) and UTOPIA (10) traffic signal control systems. However, while past experiences show that real-time traffic signal control has a potential to improve traffic operations in urban areas through their ability to employ information from traffic detectors and automatically respond to detected changes in traffic demand, significant limitations currently exist when these methods are applied to networks in which passenger cars and transit vehicles share the right of way. First, none of the existing systems considers the effects traffic of transit vehicles stopping in the right of way to board and discharge passengers. While they are stopped, these vehicles can partially or completely block a traffic lane and create a bottleneck that may result in an inefficient use of green time. Furthermore, priority of passage is also often granted to transit vehicles without considering all the potential effects that sudden traffic signal changes might have on general traffic. The Signal Priority Procedure for Optimization in Real-Time (SPPORT) model (11), was developed with the objective of addressing the above two limitations. A first effort resulted in the development of a model that could control individual intersections with a simple two-phase operation. The novelty of this model was the use of a heuristic rule-based signal operation procedure that generates timing plans on the basis of a certain number of proposed strategies. Later, Conrad et al. (12) expanded the model’s applicability to any type of intersection configuration and signal control with more than two phases. They also enhanced the rule-based signal optimization procedure and redesigned the model’s discrete-event microscopic traffic simulation module. Subsequently, an evaluation of the revised model for isolated intersections indicated its ability to provide benefits over optimal fixed time and actuated control for a range of traffic conditions (13). Recent research efforts have extended the applicability of the SPPORT model to the control of signalized intersections within coordinated networks (14-15). As a result of these enhancements, modifications have been made to the traffic demand estimation process and the rule-based signal optimization module. In particular, a new method for determining the most appropriate time for implementing signal switching decisions that explicitly considers upstream and downstream coordination impacts has been developed and introduced in the signal optimization module. This paper describes the methodology that is used to encourage progression within fully distributed network signal control systems. The scope of this paper is specifically limited to describing the methods F. Dion and B. Hellinga 3 developed for implementation within SPPORT. Consequently, limited performance evaluations of the model are presented in the paper. More exhaustive evaluations are the subject of a separate paper. In term of content, provides a brief description of the SPPORT model's system control framework. The following two sections then respectively provide descriptions of the signal optimization process and of the methods used to compute ideal phase implementation times. A fourth section then presents some evaluation results. Some conclusions and recommendations are finally made in the last section. SYSTEM CONTROL FRAMEWORK A general problem faced when designing traffic responsive signal control methods for coordinated networks is to develop methods capable of reacting to traffic variations while still maintaining a sufficient degree of coordination between adjacent intersections. Often, adding such flexibility may contradict the general objective of providing uninterrupted progression through sets of intersections. At the network level, it is usually desirable to control intersections so that vehicles released from one intersection will have a high probability of traversing downstream intersections without stopping. At the local level, the priority is usually to minimize stops and delays. Planning traffic movements on an network basis thus often involves constraining the operation of individual intersections, while providing optimum local control often implies destroying existing progression schemes to implement allow short-term tactical decisions. Therefore, the main problem is to design signal control methods that provide a reasonable balance between local and network control strategies in the presence of somewhat predictable variable demands.