摘要
针对传统遗传算法在路径规划中存在转向角度过大、转向次数过多、易陷入局部最优等问题,提出一种改进遗传算法。首先,提出一种改进种群初始化策略,即先确定一个过渡点,生成一条从起点到过渡点的路径和一条从过渡点到终点的路径,再将两条路径首尾相连成一条从起点到终点的路径,以生成优秀初始种群,提高前期搜索效率;其次,采用模拟退火算法与区域划分种群相结合的改进锦标赛选择策略,增加种群多样性,防止陷入局部最优;最后,设计一种Q-learning算法与交叉和变异相结合的策略,通过与环境交互,不断学习并优化动作选择策略以此提高算法的全局搜索能力,得到更优种群。路径规划仿真结果表明:相比传统遗传算法、改进自适应遗传算法和改进灾变遗传算法,本文所提改进遗传算法能减少路径长度和转向角度,降低转向次数,从而搜索到更优的路径。
An improved genetic algorithm is proposed to address the limitations of conventional genetic algorithms in path planning,such as excessive turning angles,redundant turns,and susceptibility to local optima.First,an upgraded population initialization strategy is proposed.This method selects a transitional point to generate a route from the starting point to the transition point and another from the transition point to the endpoint.The segments are then combined to establish a complete route,thereby forming a high-quality initial population and improving early search efficiency.Second,an enhanced tournament selection approach is utilized,which integrates simulated annealing with region-based population segmentation to increase population diversity and mitigate local optima entrapment.Finally,the Q-learning algorithm is incorporated into crossover and mutation operations,enabling the algorithm to interact with the environment,continuously learn,and optimize its action selection strategy.This integration enhances the algorithm’s global search capabilities,leading to the development of superior populations.As demonstrated by path planning simulations,the proposed enhanced genetic algorithm has the capacity to reduce path length and turning angles,decrease the number of turns,and ultimately identify more optimal pathways when compared to traditional genetic algorithms,enhanced adaptive genetic algorithms,and improved catastrophe genetic algorithms.
作者
张泽宇
王雷
蔡劲草
夏强强
ZHANG Zeyu;WANG Lei;CAI Jingcao;XIA Qiangqiang(School of Mechanical and Automotive Engineering,Anhui Polytechnic University,Wuhu 241000,China;Yangtze River Delta Hart Robot Industry Technology Research Institute,Wuhu 241000,China)
出处
《智能系统学报》
北大核心
2025年第6期1493-1504,共12页
CAAI Transactions on Intelligent Systems
基金
安徽省高校优秀拔尖人才培育项目(gxbj ZD2022023)
安徽省高校自然科学研究重点项目(2023AH050935)
安徽省机器视觉检测与感知技术重点实验室开放基金项目(KLMVI-2024-HIT-15)。
