Partially centralized hierarchical plans merging

Abstract

In multi-agent systems evolving in complex and dynamic environments, it is difficult to ensure an efficient coordination of agents plans. To deal with this issue searchers have exploited the concept of the abstraction for defining adequate plan representations, and also for providing algorithms serving to planning, coordination, and execution. By ignoring some less important details, a solution can be found more simply. In this paper we provide an adequate plan representation model, by taking advantage of hierarchical plans and the ability to reason on abstract tasks. We use the obtained model to provide some principles for implementing Partially Centralized scheme for Hierarchical Plans merging. Based on summary information and the ability to localize the interference between tasks supported by our model, the global plan is dynamically decomposed to a set of partial-global-plan that may be analysed in decentralized manner. This hybridization, between centralized and distributed coordination, can favour the interleaving of planning and execution, some part of global plan can be repaired when others are in execution. © Springer International Publishing Switzerland 2014.