引用本文:陈彦杰,王耀南,钟杭,缪志强.动态环境中服务机器人的改进型地图学习规划[J].控制理论与应用,2015,32(2):162~168.[点击复制]
CHEN Yan-jie,WANG Yao-nan,ZHONG Hang,MIAO Zhi-qiang.Improved geometrical learning planning for service robot in dynamic environment[J].Control Theory and Technology,2015,32(2):162~168.[点击复制]
动态环境中服务机器人的改进型地图学习规划
Improved geometrical learning planning for service robot in dynamic environment
摘要点击 2852  全文点击 1705  投稿时间：2014-07-06  修订日期：2014-09-12
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2015.40630
  2015,32(2):162-168
中文关键词  移动机器人  未知动态环境  路径规划  避碰
英文关键词  mobile robots  unknown dynamic environment  path planning  collision avoidance
基金项目  国家自然科学基金项目(61175075, 61433016)资助.
作者单位E-mail
陈彦杰* 湖南大学 电气与信息工程学院 chenyanjie@hnu.edu.cn 
王耀南 湖南大学 电气与信息工程学院  
钟杭 湖南大学 电气与信息工程学院  
缪志强 湖南大学 电气与信息工程学院  
中文摘要
      针对室内服务机器人进行服务工作时需要躲避碰撞和抵达目标点的功能需求, 本文提出了一种改进型地图学习路径规划算法. 在地图学习规划算法的基础上, 该算法首先约定了移动机器人的非完整性, 使规划具有更高的可行性. 然后改进了障碍物的影响方式, 令已探测到的障碍物仅对已知区域产生作用, 从而减少未知区域对路径选择的影响. 接着, 优化了地图学习算法中的随机选点策略, 即若目标点出现在探测范围内时则令目标点作为初始选取点, 解决了地图学习规划在临近目标点时收敛性不佳问题. 并设计自适应速度移动策略以进一步提高算法的收敛性能和机器人的规划效率. 最后, 仿真及实验结果表明改进型地图学习路径规划算法相比于传统地图学习算法具有更好的规划效率和目标收敛能力.
英文摘要
      To deal with the collision avoidance and target arrival for service robot when working, we propose an improved geometrical learning-based planning algorithm. Based on the geometrical learning planning algorithm, the nonholonomic constraint of mobile robot is firstly introduced to obtain higher planning feasibility. Secondly, the influence of obstacle is modified by making the detected obstacles effective only in the known area so as to reduce the influence of unknown area on path generation. Then, in order to improve the poor convergence performance of the geometrical learning planning algorithm when the robot gets close to the target, the random select point method is modified by considering the target as the first selected point when the target appears in the detected area. Moreover, an adaptive velocity moving strategy is designed to ensure the good convergence ability and high efficiency of planning. Finally, the simulation and experimental results show that the improved geometrical learning has higher planning efficiency and better convergence ability than the traditional ones.