Multilateral quality mission planning for solar-powered long-endurance UAV

Abstract

This work focuses on the development of a highly automated mission management system (MMS) for solar-powered long-endurance unmanned aerial vehicles (UAVs). The objective of the MMS is to produce a “best” plan for long endurance missions subject to the specific application's requirements and multilateral constraints, i.e. mission, energy and safety constraints. The MMS adopts the hybrid architecture of a symbolic planner based on the hierarchical task-network (HTN), working cooperatively with a Markov decision process (MDP) based policy generator to reduce the search space for a numeric path planner. The hybrid structure allows hard and soft constraints to be considered independently: the hard constraints are accounted for at each abstraction level in the task-network, while soft-constraints are considered by the policy generator. The policy generator is extended by introducing k-best policies. If the plan found by the optimal policy violates the hard constraints, a suboptimal plan will instead be selected using the suboptimal policies as ranked in the k-best policies. If multiple policies of the k-best policies find a valid plan, the operator can select the best plan by applying a Pareto rule to take into other soft constraints not considered in the determination of the k-best policies. With multilateral constraints accounted for at different hierarchical levels of the MMS, we offer more transparency to the human operator, enabling customization of the objective functions or the relaxation on hard constraints by the operator during mission execution. The MMS described in this article is especially needed for increasing autonomy of a specific fixed-wing UAV platform, namely the high altitude pseudo-satellite (HAPS). Being lightweight and fully solar-powered, the platform is practical for long-endurance surveillance and mapping missions. Due to the continuous operation over long periods, higher autonomy can yield economic and safety benefits. The MMS was tested with a lab-simulator of the HAPS.