Ying Huang,∗,∗∗ Minrui Fei,∗ and Wenju Zhou∗∗
Manipulator, trajectory planning, Pareto, MOEA/D, Tchebycheff,weight, MDP
In this study, multi-objective trajectory optimization of a manip- ulator is conducted using a multi-objective evolutionary algorithm based on decomposition (MOEA/D) under constraints. The Pareto front solution set of multi-objective trajectory planning is obtained. The Pareto front solutions after trajectory optimization of the ma- nipulator have very non-uniform distribution. In the evolutionary process of the conventional MOEA/D algorithm, the weight vectors are fixed after obtaining values in an interval. The proposed algo- rithm is enhanced after the sufficient evolution of MOEA/D. The weight vectors are adjusted online by a Markov decision process (MDP) called MOEA/D-MDP. In the end, highly reasonable weight vectors and highly uniform Pareto front solutions are obtained. The simulation shows that the function test and trajectory planning with MOEA/D-MDP improves the distribution of the two-dimensional Pareto front solutions compared with MOEA/D. Furthermore, uni- form weight vectors are not the ideal setting of MOEA/D.
Important Links:
Go Back
