基于强化蚁群算法的机器人路径规划研究  

Research on robot path planning based on reinforced ant colony optimization

作　　者：陈丹凤 雷昊 刘俊朗 何俊[1] CHEN Danfeng;LEI Hao;LIU Junlang;HE Jun(School of Mechatronic Engineering and Automation,Foshan University,Foshan 528225,China)

机构地区：[1]佛山科学技术学院机电工程与自动化学院,广东佛山528225

出　　处：《兵器装备工程学报》2023年第6期239-245,303,共8页Journal of Ordnance Equipment Engineering

摘　　要：针对蚁群算法在最优路径搜索过程中,存在参数选择过程随机性强、依赖经验值、收敛速度慢、且不同参数组合影响算法的收敛速度等问题,提出一种基于强化学习和人工势场改进的蚁群算法。首先,利用强化学习对蚁群算法的参数进行智能参数配置,即强化蚁群算法。其次,基于强化蚁群算法,引入人工势场算法的局部优化机制,针对不同维度的栅格地图进行局部路径再规划。过程中,强化蚁群算法通过对具体环境下参数的智能配置,解决参数选择过程复杂随机且依赖经验值的问题,并提升算法的收敛速度;人工势场算法的引入,通过减少局部路径的拐点数目,提升算法避障能力,实现更快更平稳的路径规划效果。结果显示,在不同维度的障碍物环境中,改进的蚁群算法都能以较快的收敛速度和较少的迭代次数搜索到最优路径,且针对高维复杂障碍物环境中的路径规划问题,改进的算法在收敛速度、迭代次数、避障能力以及路径平滑程度方面表现出更加明显的优势。本文中所提思想有望进一步扩展和推广到实际路径规划问题中,具有重要的实用意义和工程价值。Aiming at the problems of ant colony optimization(ACO)in the process of optimal path searching,such as strong randomness in parameter selection,dependence on experience values,slow convergence speed and different parameter combinations affecting the convergence speed,this paper proposes an improved ACO based on reinforcement learning(RL)and artificial potential field(APF).Firstly,RL is used to configure the parameters of the ACO intelligently,which is called the RL-ACO method.Secondly,based on the RL-ACO,the local optimization mechanism of APF is introduced to carry out local path re-planning for raster maps of different dimensions.During these processes,the RL-ACO can solve the problem of complexity and randomness of the parameter selection process,as well as the dependence on empirical values through the intelligent parameter configuration in the specific environment,and improve the convergence speed of the algorithm.In addition,the introduction of the APF algorithm promotes the obstacle avoidance ability of the algorithm by reducing the number of inflection points of the local path,and achieves faster and smoother path planning effect.The simulation results show that,in an obstacle environment of different dimensions,the improved method can search the optimal path at a faster convergence speed and with fewer iterations.For the path planning in a complex high-dimensional obstacle environment,the improved ACO algorithm demonstrates more obvious advantages in the convergence speed,iterations,obstacle avoidance ability and path smoothness.The ideas proposed in this paper are expected to be further extended and generalized to practical path planning,which has important practical significance and engineering values.

关 键 词：路径规划 强化学习 蚁群算法 参数优化 人工势场算法 

分 类 号：N39[自然科学总论]