基于分布式模型预测控制的多机器人协同编队

Multi-robot cooperative system has strong robustness and fault tolerance, which can greatly improve the overall efficiency and complete the complex tasks. At present, the multi-robot formation often adopts a centralized architecture, which relies on the central decision-making module. In particular, there are problems of insufficient scalability and low solvability when dealing with the collaborative tasks of a large number of robots. A distributed model predictive controller (DMPC) based on leader-follower method is proposed to deal with multi-robot cooperative formation tasks. The robot motion and system communication are modeled based on kinematics and graph network. The trajectory tracking and formation keeping tasks in the formation problem are decomposed, and the model predictive controllers are designed for the leader and followers, respectively. A formation matrix is designed and combined with the communication graph network to achieve consensus or formation control. Independent decision-making and parallel computing of each robot show better accuracy and scalability for the collaborative formation of a large number of robots. At the same time, the design of the controller also takes into account the change of control input, which helps to reduce energy consumption. Numerical simulation and scheme comparison are designed, and the effectiveness of the designed control strategy is verified by physical simulation experiment.