Hierarchical Multi-Agent Plans Using Model-Based Petri Net

Abstract

Agents evolving in complex and dynamic multi-agent systems need to plan their tasks and to adapt their behavior in order to react to unpredictable events. Therefore they must have plans that remain subject to continual updating, even during its execution. Although this property is important in dynamic environments, it is difficult to have a trade-off between the convergence towards the goal to be reached and the reaction to unanticipated events. To cope with this issue, it is needed to have a model allowing to represent plans less sensitive to execution contexts, and to support flexible execution. This paper aims to provide an adequate plan representation, by taking advantage of hierarchical plans and extensions to Petri net. The authors extend the Petri net to take into account the different abstraction levels of plans and the resources required by tasks. To cope with the interdependence of tasks, the authors include in the model, a synchronization mechanism between concurrent transitions. The proposed model can be used to reason on abstract levels of plans and to analyze, in a dynamic way, situations in which the execution of the plan can be performed in any ways, in some ways, or cannot be performed.