Abstract: To solve the problems of a cross-training plan for staffs in a flexible assembly cell from the point of views of humanization and economy, a multi-objective optimal cross-training plan is proposed. Average labor satisfaction and average task payment are chosen as the goals of the model, and the multi-functionality and task coverage policies are used as the constraints. To solve the multi-objective optimal model, the non-dominated sorting genetic algorithm-II based on double terminated rules and Pareto filter techniques is proposed. Simulation results show that this algorithm improves the operation efficiency and diversity of the Pareto solutions.