Auxiliary-based dual-stage constrained multiobjective evolutionary algorithm for obstacle-avoidance inverse kinematics of redundant welding manipulators

The motion planning algorithm for redundant robotic arms is a core technology for achieving efficient and precise operations in intelligent manufacturing. However, solving inverse kinematics under multiple constraints remains a major challenge in the motion planning process. To tackle this challenge, this paper reformulates the inverse kinematics problem as a constrained multiobjective optimization problem and proposes a corresponding evolutionary algorithm to solve it. Specifically, we first construct a model of the inverse kinematics problem, encompassing both objectives and constraints. The obstacle-avoidance constraint is formulated using a data-driven collision prediction method. We then introduce an auxiliary-based dual-stage constrained multiobjective evolutionary algorithm to address this problem. This algorithm subdivides the evolutionary process into an objective optimization phase and a constraint handling phase, thereby effectively balancing objectives and constraints. Besides, a global competitive swarm optimizer and a novel fitness evaluation strategy are developed in the proposed algorithm. The effectiveness of the proposed algorithm is validated by comparing it with 9 state-of-the-art constrained multiobjective evolutionary algorithms across four welding paths. The experimental results demonstrate the markedly best solving performance and middle-level time efficiency compared to peer algorithms on all welding paths. Besides, the proposed algorithm exhibits strong robustness at a data perturbation level of 5 %. This research can provide valuable insights for the field of robotic motion planning.