Abstract: To achieve efficient, safe, and energy-saving operations of tracked robots under non-flat environments, an improved non-dominated sorting genetic algorithm Ⅱ（NSGA-Ⅱ) multi-objective path planning method was presented. Firstly, comprehensively analyzing the operational requirements of tracked robots under non-flat environments, a 2.5D grid environmental model was established. Secondly, short path length, low energy consumption, and high safety were selected as sub-objective criteria for path planning. Finally, the 8-domain Manhattan collision avoidance algorithm was employed to improve the grid selection method, avoiding collisions between robots and vertices. Additionally, the NSGA-Ⅱ algorithm introduced an elite replacement strategy to expand the population size, preventing the loss of excellent path genes and accelerating algorithm convergence. Compared to the multi-objective variable neighborhood search（MOVNS) algorithm, the proposed method plans paths with an average reduction of 9.02% in path length, an average energy savings of 18.36%, and an average decrease of 7.28% in danger rate, contributing to the enhancement of path quality under non-flat environments. 

Key words: path planning, non-flat environment, energy consumption, collision-free, multi-objective optimization

摘要： 为实现非平坦环境下履带机器人路径规划高效、安全、节能等运行目标，提出了一种改进的快速精英多目标遗传算法(NSGA-Ⅱ)多目标路径规划方法。首先综合分析非平坦环境下履带机器人的作业需求，建立2.5D栅格环境模型简化复杂的环境地图；然后选取路径长度短、能量消耗低和安全性高作为机器人路径规划子目标指标；最后通过8邻域曼哈顿防碰撞算法改进栅格选择方式避免机器人与障碍物栅格顶点碰撞，并在NSGA-Ⅱ算法中引入精英替换策略扩大种群规模，防止优良路径基因流失，加速算法收敛。相较于多目标可变邻域搜索(MOVNS)算法，所提方法规划出的路径在路径长度上平均缩短9.02%、能耗上平均节省18.36%、危险率上平均降低7.28%，有助于提高非平坦环境下的路径规划质量。

关键词: 路径规划, 非平坦环境, 能量消耗, 防碰撞, 多目标优化