IBP-Laforia
1995/31:
Rapport de Recherche Laforia /
Laforia research reports
15 pages - Janvier/January 1996 -
Document en anglais.
PostScript : Ko /Kb
Titre / Title: Toward a Distributed Intelligent System for Urban Traffic Control
Abstract : We have been engaged in a project applying distributed and knowledge-based methodology for real-time, traffic-adaptive control of street and highway ramp traffic signals. We have earlier described a prototype system running in a simulated environment that proved the feasibility of the approach [Findler et al. 1992]. We have studied distributed control by a network of collaborating and learning expert systems [Findler 1991] and the dimensions of learning in real-time knowledge-based control systems [Findler 1992]. We have reported on the design features and some preliminary results of the knowledge-based approach to traffic control by a semi-autonomous decentralized system [Findler 1993, Findler et al. 1995a]. More recent efforts have concerned the integrity of inter-process/inter-node communication in distributed systems [Findler et al. 1995b] and on-line decisions about permitted/protected left turn phases [Findler et al. 1995c].
We have used two measures of effectiveness (MOEs) for the signalization at intersections: the average stopped delay per vehicle, characteristic of the level of service (LOS), and the ratio between critical flow and capacity. Our system computes a signalization regime within the optimum range of the MOEs, given the intersection geometry and the traffic flow (vehicles per hour). The system can display, store and retrieve information about the traffic flow, intersection geometry, signalization and performance of three- and four-leg intersections. In a design set-up, it can compute the least expensive geometry for an intersection, given the values of traffic flow and of an acceptable LOS.
A plausible generalization of the local MOEs has been used in optimizing traffic control in sections of an urban network of streets ("Areas of Concern"). This involves an iterative process of harmonization between intersections, leading to the best compromise for a coordinated multi-directional progression. This has been tested in the laboratory on a range of scenarios (in terms of the geometry and traffic flow) and has been found to perform well ‹ the traffic throughput could be increased on the average by about 30% while the average stopped delay remained at about the same level as it was initially when the cycle start times were set identical at each intersection.
Finally, we discuss the creation of a macroscopic model used to optimize phasing, signalization and coordinated, multi-directional progression (harmonization) as well as some other developments concerning a predictive/reactive traffic control regime.
Publications internes Laforia 1995 / Laforia research reports 1995