针对Informed-RRT(rapidly-exploring random tree)^(*)算法收敛速度慢、优化效率低和生成路径无法满足实际需求等问题,开展了基于MI-RRT^(*)(Modified Informed-RRT^(*))算法的路径规划研究,通过引入贪心采样和自适应步长的方法提高算...针对Informed-RRT(rapidly-exploring random tree)^(*)算法收敛速度慢、优化效率低和生成路径无法满足实际需求等问题,开展了基于MI-RRT^(*)(Modified Informed-RRT^(*))算法的路径规划研究,通过引入贪心采样和自适应步长的方法提高算法的收敛率,减少路径生成时间、降低内存占用;利用最小化Snap曲线优化的方法使路径平滑的同时动力也变化平缓,达到节省能量的效果,并提供实际可执行的路径。最后通过多组不同复杂度的实验环境表明,较Informed-RRT^(*)算法MI-RRT^(*)算法稳定性更高、所得规划路径平滑可执行,并且能够减少20%的迭代次数和25%的搜索时间,得出在开阔以及密集环境中MI-RRT^(*)算法较Informed-RRT^(*)和RRT^(*)算法有明显的优势。展开更多
针对快速搜索随机树(rapidly-exploring random tree,RRT)路径规划算法存在的随机性大、搜索效率低等问题,结合强化学习可根据先验知识选择策略的特点,提出了一种基于深度Q网络(deep Q-network,DQN)的改进RRT优化算法。首先设计复数域...针对快速搜索随机树(rapidly-exploring random tree,RRT)路径规划算法存在的随机性大、搜索效率低等问题,结合强化学习可根据先验知识选择策略的特点,提出了一种基于深度Q网络(deep Q-network,DQN)的改进RRT优化算法。首先设计复数域变步长的避障策略,并建立RRT算法中随机树生长的马尔科夫决策过程(Markov decision process,MDP)模型;然后将避障策略和MDP模型接入RRT-Connect算法的接口,并设计训练和路径规划的具体流程;最后在MATLAB软件平台上进行仿真实验。仿真结果表明,改进后的基于深度Q网络的RRT-Connect算法(DQN-RRT-C)在快速性和搜索效率上有显著提高。展开更多
As autonomous underwater vehicles(AUVs)merely adopt the inductive obstacle avoidance mechanism to avoid collisions with underwater obstacles,path planners for underwater robots should consider the poor search efficien...As autonomous underwater vehicles(AUVs)merely adopt the inductive obstacle avoidance mechanism to avoid collisions with underwater obstacles,path planners for underwater robots should consider the poor search efficiency and inadequate collision-avoidance ability.To overcome these problems,a specific two-player path planner based on an improved algorithm is designed.First,by combing the artificial attractive field(AAF)of artificial potential field(APF)approach with the random rapidly exploring tree(RRT)algorithm,an improved AAF-RRT algorithm with a changing attractive force proportional to the Euler distance between the point to be extended and the goal point is proposed.Second,a twolayer path planner is designed with path smoothing,which combines global planning and local planning.Finally,as verified by the simulations,the improved AAF-RRT algorithm has the strongest searching ability and the ability to cross the narrow passage among the studied three algorithms,which are the basic RRT algorithm,the common AAF-RRT algorithm,and the improved AAF-RRT algorithm.Moreover,the two-layer path planner can plan a global and optimal path for AUVs if a sudden obstacle is added to the simulation environment.展开更多
文摘针对Informed-RRT(rapidly-exploring random tree)^(*)算法收敛速度慢、优化效率低和生成路径无法满足实际需求等问题,开展了基于MI-RRT^(*)(Modified Informed-RRT^(*))算法的路径规划研究,通过引入贪心采样和自适应步长的方法提高算法的收敛率,减少路径生成时间、降低内存占用;利用最小化Snap曲线优化的方法使路径平滑的同时动力也变化平缓,达到节省能量的效果,并提供实际可执行的路径。最后通过多组不同复杂度的实验环境表明,较Informed-RRT^(*)算法MI-RRT^(*)算法稳定性更高、所得规划路径平滑可执行,并且能够减少20%的迭代次数和25%的搜索时间,得出在开阔以及密集环境中MI-RRT^(*)算法较Informed-RRT^(*)和RRT^(*)算法有明显的优势。
基金Supported by Zhejiang Key R&D Program 558 No.2021C03157the“Construction of a Leading Innovation Team”project by the Hangzhou Munic-559 ipal government,the Startup funding of New-joined PI of Westlake University with Grant No.560(041030150118)the funding support from the Westlake University and Bright Dream Joint In-561 stitute for Intelligent Robotics.
文摘As autonomous underwater vehicles(AUVs)merely adopt the inductive obstacle avoidance mechanism to avoid collisions with underwater obstacles,path planners for underwater robots should consider the poor search efficiency and inadequate collision-avoidance ability.To overcome these problems,a specific two-player path planner based on an improved algorithm is designed.First,by combing the artificial attractive field(AAF)of artificial potential field(APF)approach with the random rapidly exploring tree(RRT)algorithm,an improved AAF-RRT algorithm with a changing attractive force proportional to the Euler distance between the point to be extended and the goal point is proposed.Second,a twolayer path planner is designed with path smoothing,which combines global planning and local planning.Finally,as verified by the simulations,the improved AAF-RRT algorithm has the strongest searching ability and the ability to cross the narrow passage among the studied three algorithms,which are the basic RRT algorithm,the common AAF-RRT algorithm,and the improved AAF-RRT algorithm.Moreover,the two-layer path planner can plan a global and optimal path for AUVs if a sudden obstacle is added to the simulation environment.
