Autonomous mission planning for unmanned surface vehicles piloted by multiple specialized agents using heuristic and metaheuristic techniques

State of the art unmanned surface vehicles typically exhibit rudimentary autonomy, apart from navigation controllers. Sophisticated autopilots have enabled these vehicles to follow a path as a sequence of coordinates called waypoints, and related control tasks such as target-following, station keeping, and obstacle avoidance are well established. However, humans are typically making all the mission planning decisions. These increasingly capable platforms could offer an intelligent remote presence for the marine environment, but are used as tools with specific orders rather than as agents responsible for intelligently investigating its environment. This research attempts to increase the autonomy of unmanned surface vehicles by considering them as being controlled by multiple specialized intelligent agents, specifically, the Analyst, the Surveyor, and the Navigator. The Analyst role studies data from its environment to specify objectives. The Surveyor is responsible for conducting mission planning to efficiently meet as many objectives as possible while ensuring missions are within constraints such as time and energy limits. Missions are then executed by the Navigator. The major challenge in increasing autonomy is the high computational complexity of many of the tasks involved, such as path planning. An emphasis is placed on heuristic and metaheuristic algorithms that sacrifice optimality to make autonomy feasible. Examples of Surveyor and Analyst agents have been implemented and initial results of the techniques used to fulfill their roles are examined.