Distributed MPC based Collision Avoidance Approach for Consensus of Multiple Quadcopters

In this paper, the problem of distributed model predictive control (MPC) based collision avoidance among a team of multiple quadcopters attempting to reach consensus is investigated. A team of quadcopters trying to reach consensus or formation may collide with each other in the space if without a collision avoidance mechanism. A quadcopter receives neighbours positions and can determine a next desired position using a consensus protocol. Distributed MPC is used to develop a set of predicted relative distances along the prediction horizon. From these predicted relative distances, violations of a predetermined safety distance generates output constraints on the MPC optimization function. In literature, most strategies consider evasive action in the z-direction or only consider movement on the x-y plane. Different from past works, the proposed strategy performs collision avoidance by selecting a predetermined evasive direction in the x, y, or z-directions. This work also offers consensus in the x, y, z, ψ-directions, limits on control actions using logarithmic barrier functions and fourth-order quadcopter dynamics. The proposed algorithm was simulated for a team of quadcopters. Simulation results show that constraints on the controlled variables allows agents to converge to a consensus formation without collision.