华中科技大学学报（自然科学版）Issue(z1)：468-472,5.DOI:10.13245/j.hust.15S1111
基于低维映射和Q学习的机械臂避碰规划算法
Robot manipulator avoidance planning based on low-dimensional mapping and Q-learning
摘要
Abstract
To solve the problem of obstacle avoidance planning for humanoid robot arms,a novel algo-rithm based on the discrete low-dimensional mapping and improved Q-learning algorithm was pro-posed.The problem of obstacle avoidance in three-dimensional spaces was converted to a similar prob-lem of mobile robots obstacle avoidance using discrete low-dimensional mapping.A discrete envelope method was implemented to conduct the collision detection,which improved the efficiency of collision detection.A decision tree produced by look-ahead searching was supplemented to the typical Q-learn-ing algorithm for searching possible solutions,which avoided the defects of traditional Q-learning al-gorithms that were prone to have no solutions.Afterwards,the typical Q-learning algorithm was ap-plied to improve the available solution.Finally,the feasibility of the proposed algorithm was demon-strated by some comparative simulation tests.关键词
机器学习/Q 学习/机械臂避碰/离散化低维映射/前向树Key words
machine learning/Q-learning/robot avoidance planning/discrete low-dimensional map-ping/look-ahead decision tree分类
信息技术与安全科学引用本文复制引用
钱堃,贾凯,宋潇..基于低维映射和Q学习的机械臂避碰规划算法[J].华中科技大学学报（自然科学版）,2015,(z1):468-472,5.基金项目
江苏省科技成果转化项目（BA2014044）；东南大学 SRTP 项目（1410286049）． （）
