Optimization of Decentralized Task Assignment for Heterogeneous UAVs

In this paper, the optimization of a decentralized task assignment strategy for heterogeneous UAVs in a probabilistic engagement scenario is investigated. In the engagement scenario, each UAV selects its targets by employing the consensus-based bundle algorithm (CBBA). This paper uses a scoring matrix to reflect heterogeneity among the UAVs and targets with different capabilities. Therefore, a performance improvement of CBBA is closely connected with the scoring matrix and it should be optimally selected. The values of scoring matrix can be obtained by employing an episodic parameter optimization (EPO). The EPO algorithm is performed during the numerous repeated simulation runs of the engagement and the reward of each episode is updated using reinforcement learning. The candidate scoring matrices are selected by using particle swarm optimization. The optimization results show that the team survivability of the UAVs is increased after performing the EPO algorithm and the values of the optimized score matrix are also optimally selected.