Distributed model predictive control for unmanned aerial vehicles

Abstract

In this article a distributed model predictive control scheme, for the cooperative motion control of Unmanned Aerial Vehicles (UAVs) is being presented. The UAVs are modeled by a 6-DOF nonlinear kinematic model. Two different control architectures: a centralized and a distributed MPC, are studied and evaluated in simulation experiments. In the centralized approach, one central MPC controller is responsible for the movement coordination of all the UAVs, while in the distributed approach each aerial vehicle plans only for its own actions, while the objective function is coupled with the behavior of the rest of the team members and the constraints are decoupled. In this approach, each agent only shares the future position of itself with the other agents to avoid collisions. For reducing the computation time and complexity, only one step ahead prediction in the corresponding MPC schemes have been considered without a loss of generality. Finally, the efficiency of the overall suggested decentralized MPC scheme, as well as it comparison with the centralized approach, is being evaluated through the utilization of multiple simulation scenarios.