Interleaving Deliberative and Reactive Planning in Dynamic Multi-Agent Domains

Reactive planning, consisting of pre-defined sensor action rules, is well suited to effectively respond to dynamic changes in real-time environments, However, it is in general challenging to strategically reason about long or short-term objectives using reactive planning. Therefore, ideally, deliberative and reactive planning should be integrated, In this paper, we introduce an adaptive interleaving of deliberative and reactive planning as our approach for dealing with real-time dynamic environments. Two main aspects are responsible for the success of the approach, First, the deliberative planner uses depth-bounded forward chaining guided by goal-based heuristics. Second, the real-time state space is discretized as a function of the average time that the deliberative planner needs to generate a plan. This ensures that the state, as seen by the deliberative planner, does not change in average while the plan is being generated. When a plan fails or a new plan is needed, the reactive planner takes over. We extend our approach to multi-agent real-time domains, where the need for collaborative deliberative planning is particularly needed. We implemented and demonstrate our integration using the Prodigy deliberative planner and the RoboCup soccer simulator server.