Bull, L; Sha'Aban, J; Tomlinson, A; Addison, JD; Heydecker, BG; (2004) Towards distributed adaptive control for road traffic junction signals using learning classifier systems. In: Bull, L, (ed.) Applications of learning classifier systems. (pp. 279-299). Springer: New York.

Preview

PDF 2004_29.pdf Download (111kB)

Abstract

This chapter considers an approach to distributed traffic responsive signal controlusing Learning Classifier Systems (Holland, 1976). The intention is to accommodaterealistic kinds of detector data and wide ranges of candidate performance criteria fortraffic management in a fully flexible manner. The approach to achieving this is to useevolutionary computing (eg Holland, 1975) and reinforcement learning (eg Sutton andBarto, 1998) with performance fed back from microscopic traffic simulations: thisapproach has the advantage that it is not specific to any particular objective or form ofprimary data. The purpose of this work is to develop an approach to distributedoptimisation that can achieve good traffic performance flexibly according to any on arange of possible criteria using data from existing traffic detectors. Here each junctionin a road network is controlled by a Learning Classifier System using only locallyavailable input and performance data; a multi-agent approach is proposed.Learning Classifier Systems (LCS) can be used for optimisation in a way thatoffers substantial promise for application in traffic-responsive signal control systemswhere the way in which the control responds to variations in traffic flows can beadapted according to measured conditions. This is important in order to achieve trafficcontrol that is sufficiently flexible to respond rapidly when traffic conditions change ina fundamental way, as occurs at the start of a peak period, without being undulysensitive to short-term variations in flow. The expectation is that this will be possibleby their use of both reinforcement learning and evolutionary computing techniques.Furthermore, they offer the automated rule development of neural networks togetherwith the transparency of production system rules.The importance of this approach for traffic control is that it offers a means bywhich signal control strategies can be developed directly according to theirperformance, evaluated using detailed microscopic simulation as opposed to thatestimated from formulae that have been adopted on grounds of analytical convenience.This closed-loop approach to development of control strategies offers severaladvantages over the use of traditional explicit optimisation formulations. Theseinclude flexibility in respect of objectives so that multiple and varying needs can beaccommodated, ability to use various different kinds of detector data according to theiravailability, and freedom from dependence on a single explicit evaluation formula thatis intended to embody the whole of a traffic model. This final point has been found tobe especially important in recent research work where certain fine details of themodels used have been found to have an unexpectedly strong influence onperformance.

Download activity - last month

Export as CSVJSONXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as JSONXMLCSV

Archive Staff Only

View Item