Research on system planning for multi-robot welding based on ARES-MOEA

Efficient planning of multi-robot welding systems is essential for improving production efficiency and reducing robot motion burden in automotive manufacturing. However, existing studies rarely consider task assignment, welding sequence planning, and welding pose selection simultaneously. To address this issue, a tri-objective joint optimization model is established at the task planning level for multi-robot welding systems. The model considers the estimated production cycle time, potential task transition cost, and weighted joint motion cost of the two robots as optimization objectives, while a three-layer chromosome encoding scheme is developed to represent the coupled decision variables within a unified framework. To improve the search performance in the resulting complex search space, an Adaptive Reference Energy-driven Selection Multi-Objective Evolutionary Algorithm is further proposed, integrating bidding-based population initialization, a local search operator, an adaptive reference point mechanism, and energy-driven environmental selection. The proposed method is evaluated through standard multi-objective benchmark problems, ablation studies, and a real-world vehicle floor panel welding case. Experimental results demonstrate that ARES-MOEA achieves competitive convergence and solution diversity, while the ablation studies verify the effectiveness of its individual components. For real-world welding tasks of different scales, the proposed method consistently produces high-quality planning solutions with favorable overall performance, demonstrating its effectiveness for practical multi-robot welding task planning.