Robust Decentralized Task Assignment for Cooperative UAVs

This paper investigates the problem of decentralized task assignment for a fleet of cooperative UAVs. It extends the analysis of a previously proposed algorithm to consider the performance with different communication network topologies. The results show that the second communication step introduced during the planning phase of the new algorithm is crucial for sparse networks because the convergence rate of the information consensus algorithms can be quite slow. Further analysis shows that the selection of the candidate plans communicated during this planning phase has a significant impact on the performance of the overall algorithm. A comparison of the performance and computation of four selection approaches clearly shows the importance of correctly accounting for the potential actions of the other UAVs, even though that tends to be more computationally expensive. A modification of the original candidate plan selection algorithm is also presented, which further improves the overall performance by increasing the robustness to inconsistencies in the information across the team.