An analysis of traffic-flow stability in a microscopic heterogeneous network

摘要

Microscopic traffic-flow networks are typically designed to simulate vehicle acceleration behaviour using a single mathematical model. The concept of stability is a major objective of microscopic driver models for traffic simulations. Stable flow is traffic flow that does not fluctuate unaccountably, and changes in flow do not unreasonably magnify downstream, thereby reflecting real-world driver behaviour. However, the stability of driver models is typically evaluated in isolation, with the single model applied to all vehicles and road sections in the traffic network. Yet as different models will be more effective in different situations, it would be desirable to mix multiple models within the one traffic network. Heterogeneous approaches that mix microscopic and macroscopic models exist, but mixing different types of microscopic driver models has been largely overlooked and no analysis of the consequences on stability has been made. Thus this paper investigates the stability characteristics of such microscopic heterogeneous networks, mixing the well-known Intelligent Driver Model (IDM), a continuous-space car-following model, with a discrete-space cellular automata model. To this end, a flaw in the stability of the IDM at the speed limit is identified and corrected. Subsequently it is shown that model switch-over points will experience instabilities despite the stability of the individual driver models and, although careful choice of parameters can reduce the problem, it cannot be completely eliminated in practical road networks. However, the instability produces a signature 'fingerprint' effect on traffic density, and this fingerprint is readily identifiable using simple measures of traffic flow even in realistic road networks.