Multi-objective Trajectory Optimization of the 2-Redundancy Planar Feeding Manipulator Based on Pseudo-Attractor and Radial Basis Function Neural Network

The establishment and solution of the inverse kinematic model is the key to improve the efficiency of trajectory optimization. To improve the trajectory smoothness and reduce energy consumption of multi-degree-of-freedom (MDOF) robots, this article presents the time-, jitter-, and energy-optimal trajectory optimization method based on pseudo-attractor and radial basis function neural network. Based on the geometric method, the forward kinematic model of MDOF robots is firstly established. The diversity of inverse kinematic solutions is reduced by determining redundant joints. Combined with the attractor theory, the time-adaptive allocation strategy can automatically endow time information with path points. On this basis, the 7-time polynomial interpolation method is used to fit discrete trajectory points and generate the initial trajectory without singularity points. Affected by the pseudo-attractor, radial basis function neural network is transformed into the improved radial basis function neural network (I-RBFNN) to optimize the initial trajectory. The 2-redundancy planar feeding manipulator (2-RPFM) is introduced to verify the effectiveness of the proposed method. Experiment and simulation results show that the proposed method is available in generating high-performance trajectories, which is beneficial to improve the production efficiency of the auto-body-out-panel stamping line.