Multi-Manipulator Motion Planning based on Gaussian Process with Probabilistic Inference

In this paper, we study the problem of using Gaussian process motion planning to plan the motion of a multi-manipulator. The DH (Denavit Hartenberg) method is used to establish the kinematic model of the multi-manipulator system. The GPMP2 algorithm is extended to the case of multi-manipulators, and constraints are added to the factor graph to avoid mutual collision between arms and limit the range, speed, and acceleration of joints. Because the GPMP2 algorithm does not include acceleration prior, we convert the limitation of acceleration into a speed limitation. Depending on the actual conditions of the different arms in the system, different constraints are imposed, which can better realize the cooperation between multi-manipulators. The optimized method is more reliable in solving the actual motion planning problem. Simulation results demonstrate the high effectiveness and feasibility of this method.