Dynamic Task Allocation for Heterogeneous UAVs Based on Task Evaluation Feedback

Task allocation for heterogeneous UAVs in a dynamic battlefield environment is one of the critical techniques for rapid combat capability generation of UAVs. A dynamic task allocation method based on task execution feedback is proposed to address the challenges of task demand changes and potential feasibility issues caused by battlefield uncertainty which have not been well-solved by existing studies. By designing a task evaluation feedback module, real-time monitoring of the UAVs’ task performance is achieved, and an event triggered mechanism is utilized to respond to changing task requirements in a dynamic environment. Heterogeneous UAVs allocation is implemented based on the deep Q-learning network. By the utilizing the decoupling of task allocation and execution as well as a designed action filtering mechanism, the difficulty of training convergence caused by sparse rewards in traditional reinforcement learning is overcome, which forms a dynamic and fast real-time task response scheme. The simulation results show that the proposed method can effectively deal with task failures caused by uncertain battlefield environments, and ultimately achieve a better task completion rate than existing methods.