基于置信度上界的移动机器人信息路径规划方法
作者:
作者单位:

1.天津大学;2.北京空间飞行器总体设计部

作者简介:

通讯作者:

中图分类号:

TP242

基金项目:

国家自然科学基金项目(面上项目);国家重点研发计划项目子课题;国家自然基金面上项目


An Informative Path Planning Approach for Mobile Robots Based on Confidence Upper Bound Algorithm
Author:
Affiliation:

Tianjin University

Fund Project:

The National Natural Science Foundation of China (General Program);The National Key Research and Development Program of China, Sub-project

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    路径规划是移动机器人未知环境探索的关键问题.路径点的合理规划对提高环境探索的效率和环境场预测的准确性至关重要.本文基于强化学习范式,提出一种适用于静态环境场探索的移动机器人在线信息路径规划方法.针对基于模型训练算法计算成本高的问题,通过机器人与环境的交互作用,采用动作价值评估的方法来学习所获取的环境场历史信息,提高机器人实时规划能力.为了提高环境预测准确性,引入基于置信度上界的动作选择方法来平衡探索未知区域与利用已有信息,鼓励机器人向更多未知区域进行全场特征探索,同时避免了因探索区域有限而陷入局部极值.仿真实验中,环境场分别采用高斯分布和Ackley函数模型.结果表明,本文算法能够实现机器人环境探索路径点的在线决策,准确有效地捕捉全场和局部环境特征.

    Abstract:

    Path planning for mobile robots is paramount in unknown environment exploration. Exploration efficiency and prediction accuracy largely depend on appropriate waypoint decision. In this paper, an informative path planning approach is proposed based on the reinforcement learning paradigm for static environment exploration. In contrast to model based algorithms, no assumption is presumed for the environmental features. The computational cost is reduced and the online planning capability is enhanced by evaluating the values of actions through robot’s interaction with the environment. To improve the prediction accuracy, an action selection algorithm based on the Upper Confidence Bound (UCB) is utilized to balance exploration and utilization. Exploration in unknown areas is encouraged, which also potentially avoid stucking into local extrema. Numerical simulations have been performed on environments that are modeled with Gaussian distribution and Ackley function respectively. Results show that the characteristics of the entire environment field can be effectively captured using the proposed path planning approach.

    参考文献
    相似文献
    引证文献
引用本文
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2021-07-03
  • 最后修改日期:2021-10-30
  • 录用日期:2021-11-10
  • 在线发布日期: 2021-12-01
  • 出版日期: