Development of advanced multi-agent control strategies for multi-class traffic networks

Project members: M. van den Berg, B. De Schutter, J. Hellendoorn

Sponsored by: NWO-CONNEKT

This project is part of a larger project ``Advanced multi-agent control and information for integrated multiclass traffic networks'' (AMICI), and aims at developing an innovative control theory specifically suited to the coordinated control of heterogeneous traffic flow. General control objectives are to optimize network efficiency, to minimize safety-critical traffic conditions, and to minimize societal impacts, by providing class-specific travel information and traffic control. We combine existing control design techniques from the fields of distributed and hierarchical control, supervisory control, model-based control, hybrid systems control and multi-agent control and apply them to the traffic control setting sketched above. These techniques will be used to obtain a structured control design method and adapt them to the typical characteristics of a traffic system.

Figure 27: Multi-agent control framework.

Figure 27 illustrates the multi-agent control framework used in this project. The traffic network is divided in several possibly overlapping regions, each of which is controlled by a local control agent. The agents cooperate and coordinate their actions so as to contribute to the reliable and robust operation and the performance of the entire system. Furthermore, we have one or more higher ``supervisory'' control levels.

In this project we also focus on optimization of the interaction between the urban network and the motorways, by dividing the networks into subregions. Besides being class-specific, integrated and coordinated, another characteristic is that the developed traffic control systems will be anticipatory (i.e., adaptive and predictive).