Long-Term Utility Optimization-based Mission Assignment and Trajectory Planning for Multi-UAV Cooperation Scenarios

Unmanned aerial vehicles (UAVs) have attracted widespread attentions for the applications in various fields leveraging their high maneuverability, ease of deployment, and low cost. Trajectory optimization and mission allocation of UAVs are among the most critical issues for achieving flexible and efficient mission execution. In this paper, we study the problem of trajectory planning and mission assignment in a mission execute scenario of multiple UAVs. Considering the reward obtained from mission execution, as well as the energy and resource consumption due to mission execution, we define system utility function and formulate the joint trajectory planning and mission assignment problem as a constrained system long-term utility function maximization problem. To solve the formulated problem, we first model the flight trajectory planing problem of the UAVs as a decentralized partially observable Markov decision process (Dec-POMDP) and propose a multi-agent double deep Q network (DDQN)-based trajectory planning algorithm. Given specific system state and action within the DDQN framework, we then design a greedy method-based mission assignment and execution ratio determination strategy. Simulation results demonstrate the effectiveness and superiority of the proposed algorithms.