Robust UAV coordination for target tracking using output-feedback model predictive control with moving horizon estimation

Abstract

We consider the control of two UAVs tracking an evasive moving ground vehicle. The UAVs are small fixed-wing aircraft equipped with gimbaled cameras and must coordinate their control actions so that at least one UAV is always close to the target. The control actions of the UAVs are computed based on noisy measurements of the UAVs' current state and vision-based measurements of the target's position corrupted by state-dependent noise. We propose a novel approach for solving this problem in which the state estimate and optimal control are computed simultaneously online using model predictive control with moving horizon estimation. The efficacy of this approach is demonstrated in simulation results using realistic vision-based measurements of the target's position. These results show that while using noisy, partial information about the target state, the UAVs are able to coordinate their distances to the target in the ideal case of constant target velocity as well as perform robustly when the target becomes evasive.