A network-wide assessment of local signal control policies' performance in practical implementations

Abstract

This paper deals with the critical problem of solving both the traffic control and assignment problems simultaneously. The main purpose of this work is to assess two well-known local control policies - Equisaturation and P0 - in terms of network-wide performances. While local control policies are, indeed, key building blocks to network-wide approaches, little effort has been done to evaluate their performances for real-life scenarios. While Equisaturation has been regarded as the most popular local policy, it is also widely accepted that this policy is more ‘fair’ than ‘efficient’. The main contribution of this paper is to test the properties of these two approaches on a real sized network. The local control policies are employed so to generate a starting point from which further optimization is then performed through a network-wide approach - explicitly maximizing network throughput. In order to ensure a reliable performance evaluation, different random initial points are tested. The contribution is twofold: first, authors compare the two control policies and assess their performances in approximating network-wide approaches. Subsequently, a centralized optimization is performed, to evaluate the impact of generating ad hoc initial points when optimizing multiple intersections.