基于分层边界与可视图的移动机器人自主探索算法  

An autonomous exploration algorithm of mobile robots based on hierarchical frontier and visibility graph

作　　者：曹路阳 周乐来 戴骁蒙 张文鹤 李贻斌[1] CAO Lu-yang;ZHOU Le-lai;DAI Xiao-meng;ZHANG Wen-he;LI Yi-bin(School of Control Science and Engineering,Shandong University,Jinan 250061,China;China Railway Construction Bridge Engineering Bureau Group Construction Assembly Technology Co.,Ltd,Tianjin 301699,China)

机构地区：[1]山东大学控制科学与工程学院,济南250061 [2]中铁建大桥工程局集团建筑装配科技有限公司,天津301699

出　　处：《控制与决策》2025年第4期1207-1216,共10页Control and Decision

基　　金：国家自然科学基金项目(62373221);山东省杰出青年基金项目(ZR2022JQ28);中铁建大桥工程局集团科技项目(DQJ-2022-A03);天津市科技计划项目(23ZGCXQY00030)。

摘　　要：为了减少机器人在探索过程中容易忽视局部狭小区域、路径重复度高、探索效率低下的问题,提出一种基于分层边界与可视图的自主探索算法.首先,根据三维地图中状态变化的体素,实时提取局部边界并增量构建全局边界,对边界聚类得到候选目标点;其次,基于增量更新的可视图对候选目标点进行综合指标的评价,采用一种指数衰减形式的评估函数;再次,将可视图与D*Lite算法结合,基于动态规划的思想,引导机器人快速完成对未知环境的探索,避免重复路径;最后,在不同环境下进行仿真实验,通过数据验证所提出方法在移动距离、运行时间、探索效率方面都优于次优视图规划器(NBVP)、基于图的探索规划器2 (GBP2)和双阶段视点规划器(DSVP)算法.结果表明,该算法可以有效解决机器人在探索时忽视局部狭小区域、路径重复度高的问题,提高了机器人自主探索的效率.In order to reduce the problems of the mobile robot easily overlooking local narrow areas,high path redundancy,and low exploration efficiency during the exploration process,an autonomous exploration algorithm based on hierarchical frontier and visibility graphs is proposed.Firstly,the local frontier is extracted in real-time based on the state-changed voxels in the OctoMap and the global frontier is incrementally constructed,clustering the frontier to obtain candidate target points.Secondly,candidate target points are evaluated using a comprehensive index based on an incrementally updated visibility graph,by employing an evaluation function in the form of exponential decay.Thirdly,the visibility graph is integrated with the D*Lite algorithm,guided by the principles of dynamic programming,to facilitate rapid completion of exploration of unknown environments by the robot and avoid redundant paths.Finally,simulation experiments conducted in different environments show that the algorithm outperforms next-best-view planner(NBVP),graph-based planner2(GBP2)and dual-stage viewpointplanner(DSVP)algorithms in terms of distance traveled,runtime,and exploration efficiency.These results indicate that the algorithm effectively addresses the problems of overlooking local narrow areas and high path redundancy,thereby improving the efficiency of robot autonomous exploration.

关 键 词：自主探索 未知环境 边界检测 代价评价 可视图 机器人操作系统 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置]