An Evaluation of Adaptive Decentralised Nature Inspired Traffic Light Control

Road traffic is an increasingly large problem which often blights urban areas. Continually building additional infrastructure is not only costly but also often unwelcome if not impossible. The cause of traffic jams however are not always, for example, a direct result of accidents or insufficient road capacity. Traffic jams can often be attributed to the interactions of individual drivers and are therefore an emergent phenomena. Controlling traffic intelligently is therefore a difficult problem but an essential part of using existing infrastructure effectively. Many species of social insects, such as termites and honeybees, exhibit self-organisation. Colonies of such social insects are decentralised systems that are capable of demonstrating efficiency, robustness and flexibility. The response threshold algorithm is capable of explaining how, without centralised control, individual social insects can specialise upon specific tasks whilst as a whole producing coordinated behaviour to meet the needs of the colony. Furthermore, this algorithm has been applied to resource allocation problems within multi-agent systems. Results from these studies show that the algorithm causes agents to specialise effectively within both static and dynamic environments. The application of the response threshold algorithm upon a traffic light control problem contrasts with the more traditional centralised approaches but has several potential benefits. The implementation of traffic control systems based upon the response threshold is discussed. In addition, significant performance testing is included along with detailed potential future work and conclusions.