Safe Formation Control of Multi-Agent Systems via Control Barrier Function and Hybrid Disturbance Rejection

This paper proposes a hybrid formation control framework for multi-agent systems to ensure hard safety constraints subject to unknown disturbances. While extended state observers (ESOs) actively compensate for lumped disturbances, their residual estimation errors often lead to constraint violations and compromise system safety. To address this, we propose a decoupled architecture, employing ESO-based reference control signals for formation tracking and distributed modified control barrier function (DMCBF) based safety filters. By integrating the DMCBF into a quadratic program (QP), it provides passive robustness against bounded estimation errors, strictly enforcing inter-agent and agent-obstacle collision avoidance. It demonstrates that the safety constraints and input-to-state stability (ISS) are guaranteed. Finally, experiments on a Crazyflie quadrotor swarm under wind disturbances validate the practical effectiveness of the proposed control framework.