Fast task allocation for heterogeneous unmanned aerial vehicles through reinforcement learning

摘要

A task allocation problem for heterogeneous unmanned aerial vehicles (UAVs) in the presence of environment uncertainty is studied in this paper. Generally, the process of finding efficient allocation scheme can be computationally prohibitive. This work presents a Q-learning based fast task allocation (FTA) algorithm through neural network approximation and prioritized experience replay, which effectively offloads the online computation to an offline learning procedure. Specifically, the proposed approach develops a Q network that encodes the allocation rules. The Q network not only considers the effect of environment uncertainty, but also is capable of handling total different tasks. Comparison simulations are provided to show the efficiency of the proposed algorithm. (C) 2019 Elsevier Masson SAS. All rights reserved.