Multi-agent Strategy for Marine Applications via Temporal Planning

Abstract

This paper proposes a novel strategy for the implementation of complex missions in an underwater domain, based on the combination of a Goal Allocation (GA) strategy and Temporal Planning (TP). We define a new highly constrained planning domain, motivated by the autonomous supervision of an offshore wind farm scenario. The domain introduces actions which enhance the execution of long-term missions, support coordinated actions between surface and underwater vehicles and reduce the risks of failures. The ROSPlan framework, a realistic AUV, and Autonomous Surface Vehicle (ASV) simulators are integrated to implement effective missions, and plan execution is analysed using ROS and Gazebo. Experiments have shown a dynamic refuelling point method enhances long-term missions with small robot fleets. The approach also contributes to an increase in plan quality while relaxing complexity in the planning process, leading to an improvement in the planning time and makespan results for multi-agent problems.