Centralized cooperative planning for dynamic multi-agent planar manipulation

This paper studies the coordination problem for dynamic multi-agent planar manipulation, and proposed a novel centralized cooperative planning method based on backstepping design and quadratic programming. The objective of coordination is to plan interaction forces between agents and object, such that the object can follow a given trajectory, and the forces satisfy the predefined performance index. In this paper, the coordination problem is solved hierarchically in two levels. In the lower control level, a generalized force input is designed using the backstepping technique, with which the agents can control the object tracking a given trajectory. In the higher coordination level, the distribution of forces between agents is discussed in a quadratic programming framework, and the optimal force distribution is found by solving a quadratic programming problem. Simulations are carried out for two-agent and three-agent manipulations, and the results obtained demonstrate the effectiveness of the proposed coordination method.