The Network Signal Design Problem for Long-Range Travel Forecasting

Abstract

The network signal design problem (NSDP) seeks the optimal deployment of traffic signals in a growing urban area. This paper is especially concerned with how signals may be optimally deployed over a very long period of time for the purpose of creating realistic networks for travel forecasting. The NSDP is very difficult to solve for long-range problems because of the large number of possible solutions, the high cost of evaluating the merits of just a single solution, and the complexities of how signal delay affects traffic patterns and how traffic patterns affect signal delay. The paper describes the NSDP, introduces a reasonable set of simplifications based on transportation planning and traffic engineering practice, describes experiences with a possible heuristic algorithm for problem solution, and contrasts this method with current planning practice and other research. The long-range algorithm embeds a “strategic” algorithm for finding an optimal deployment for a single time period with constant travel demands. The strategic algorithm draws upon two well-known techniques of combinatorial optimization: a greedy constructive search coupled with a restricted neighborhood search. The strategic algorithm was able to find exact solutions on a small test network with eight stop-controlled intersections. The long-range algorithm is demonstrated on a full-sized planning network with about 380 stop-controlled intersections that could be signalized.