A Visual Predictive Control Framework for Robust and Constrained Multi-Agent Formation Control

In this study, the problem of leader-follower position-based formation control is considered. Each agent in the multi-agent network is equipped with a perspective camera (assumed to be pinhole-type) and is servoed visually. A depth-based visual predictive controller is proposed. The framework optimizes the planned trajectory with a prediction horizon, while taking image- and physical-space constraints into account. Furthermore, the presented control scheme provides robustness against the camera occlusion, modeling errors and uncertainties. The performance of the proposed tracking algorithm is validated via numerous simulations.