DAWN: Dynamic Task Planning of Multi-UAV With Two-Layer Optimization Mechanism in Uncertain Environments
Daqian Liu, Bowen Fei, Weidong Bao, Xiaomin Zhu, Xiaoqing Li
Abstract
UAV cooperative formation provides rescue and material delivery for the industrial Internet of Things (IIoT). To solve issues, such as low material distribution efficiency and poor mobility during disaster rescue, we propose a two-layer optimization mechanism-based multiple UAV dynamic task planning method (DAWN), which can cope with the problem of the global communication link unreachable caused by disasters. Specifically, we consider the global task allocation as a dynamic vehicle routing problem (VRP) and use deep reinforcement learning (DRL) to solve it so as to minimize the global flight path and energy consumption. Second, based on the current communication structure, we establish a local path planning approach based on the trust network that maximizes the regional coverage rate while minimizing the flight paths. On the basis of these two layers, an UAV formation dynamic task planning approach is realized. Experimental results prove that the proposed DAWN can obtain the optimal flight paths and achieve higher energy efficiency while providing reasonable region coverage to discover more potential tasks.