Multi-Agent Planning under Complex Constraints for Deep-Space Probes Group

With the development of launch technology and the miniaturization of probes, multi-probe systems can be used in space missions, and the demand for on-board planning is rising. Multi-probe planning needs tasks to be allocated properly to increase the total task profit and satisfy the complex time and resource constraints of probes. We propose to solve this problem using a multi-agent system, taking advantage of its negotiation strategy in task allocation to increase the global task profit. And we add complex constraint handling ability to agents, to deal with the more realistic and detailed constraints models in practical engineering. Before planning starts, a dynamic distributed multi-probe group formation pattern is designed to cope with the unstable communication condition between probes. In the multi-agent system, two types of agents are designed: execution agents represent task accomplish abilities of probes and target agents represent target requirements in the task. A novel three-phase multi-agent negotiation planning method is proposed, target agents ask for high-profit actions and execution agents respond depend on constraints and optimal strategy. The last phase of planning satisfies resource constraints through an action-bind strategy, which can handle constraints between overlapped resource consumption and resource generation actions. The experimental results show that the proposed multi-agent planning method can obtain rational plans satisfying complex constraints and increases the task profit compared to the traditional greedy search method.