<p>With the development of Industry 4.0 technologies, collaborative robots are increasingly utilized to assist workers to assemble tasks or complete the assembly tasks solely in assembly lines. In such type oflines, human health is an important factor which influences assembly line efficiency. In the last few years, study on ergonomics in assembly lines has gained more attentions. This study investigates the assembly line balancing problem with collaborative robots considering ergonomics. A mathematical model is developed to minimize cycle time and the ergonomics risks difference among the human workers. Meanwhile, a multi-objective restarted simulated annealing algorithm is proposed to solve this problem effectively. The developed algorithm proposes three vectors for encoding, where task assignment vector tackles task allocation sub-problem, process alternative vector tackles process alternative allocation sub-problem and task process alternative selection vector tackles process alternative selection sub-problem. Basedon the information of the three vectors, a decoding procedure is employed to achieve one feasible solution and achieve the ergonomics risks of stations. In addition, the proposed algorithm includes two improvements: (1) a new multi-objective acceptance criterion to accept new solution; and (2) a restart mechanism to replace the current solution with one selected solution from the permanent Pareto solutions based on the modified crowding distance. Comparative study demonstrates that the proposed acceptance criterion and restart mechanism outperform the original version. To evaluate the performance of the proposed method, it is compared with four other multi-objective algorithms. Computational study demonstrates that the proposed method achieves promising performance by outperforming the compared methods.</p>