Abstract: Aiming at the problem of low task completion caused by task completion time constraints in the scenario where tasks are randomly assigned, a distributed and dynamic multi-UAV task allocation method based on deep reinforcement learning is proposed. The method uses the interaction between UAVs to quantify the time constraints, task size, task priority and other characteristics of the tasks being performed and new tasks in real time. At the same time, new task priority features are generated dynamically during task execution. Then, the task features after UAV interaction are regarded as the global task shared by UAV, and constantly updated to form a dynamic decision-making basis. Finally, according to the real-time task completion and the time constraints, the behavioral decision-making of the UAV is made based on the deep reinforcement learning, so as to improve the task completion by achieving the dynamic allocation of new tasks and ongoing tasks. This behavioral decision is to realize the dynamic assignment of tasks to improve the task completion. Simulation results show that this method can improve the overall task completion of the system under time constraints.

Key words: UAV, task allocation, deep reinforcement learning, dynamics, task completion