第16卷第1期 智能系统学报 Vol.16 No.1 2021年1月 CAAI Transactions on Intelligent Systems Jan.2021 D0:10.11992/tis.202007025 室外未知环境下的AGV地貌主动探索感知 张威'，葛泉波2，刘华平3，孙富春3 (1.上海海事大学物流工程学院，上海201306,2.同济大学电子与信息工程学院，上海201804,3.清华大学计算 机科学与技术系，北京100084) 摘要：智能机器人对复杂地貌环境的识别一直是机器人应用领域研究的前沿问题，移动机器人在不同的地貌 上采取的运动方式并非一成不变，所以选择的运动方式对于迅速准确识别所处地貌的类型至关重要。针对该 问题本文提出了一种基于贝叶斯框架的主动感知探索方法，使移动机器人能够主动探索有兴趣的运动方式并 且感知识别和运动之间的匹配关系，可以优化在地貌识别之中的模糊不确定性；为了进一步验证实验的可靠 性，还使用了被动感知策略来比较和分析不同策略之间的差异。实验结果表明：主动感知方法能够规划出有效 的地貌识别动作序列，能够引导移动机器人主动感知目标地貌，该框架对于室外未知环境下主动感知后的地貌 识别效果优于被动感知。 关键词：移动机器人；运动方式；贝叶斯框架：主动感知；被动感知；地貌识别：振动数据；室外地貌 中图分类号：TP391文献标志码：A文章编号：1673-4785(2021)01-0152-10 中文引用格式：张威，葛泉波，刘华平，等.室外未知环境下的AGV地貌主动探索感知.智能系统学报，2021,16(1)： 152-161. 英文引用格式：ZHANG Wei,,GE Quanbo,LIU Huaping,.etal.AGV active landform exploration and perception in an unknown outdoor environmentJ].CAAI transactions on intelligent systems,2021,16(1):152-161. AGV active landform exploration and perception in an unknown outdoor environment ZHANG Wei,GE Quanbo',LIU Huaping',SUN Fuchun' (1.Logistics Engineering College,Shanghai Maritime University,Shanghai 201306,China;2.School of Electronics and Information Engineering,Tongji University,Shanghai 201804,China;3.Department of Computer Science and Technology,Tsinghua University, Beijing 100084,China) Abstract:The recognition of a complex landform environment by an intelligent robot has been the frontier problem of research in the field of robotics applications.The motion modes adopted by mobile robots differ between landforms,so the selected motion mode is crucial for quickly and accurately identifying the type of landform.To solve this problem, an active perception exploration method is proposed in this paper based on a Bayesian framework.It enables mobile ro- bots to actively explore interesting motion modes and recognize the matching relationship between landform and move- ment.It can optimize the fuzzy uncertainty in landform recognition.To further verify the reliability of the experiment, we also use a passive recognition strategy to compare and analyze the differences between different strategies.The ex- perimental results show that the active perception method can plan effective landform recognition action sequences and guide mobile robots to actively perceive the target landform.The landform recognition effect of active perception is bet- ter than that of passive perception in an unknown outdoor environment. Keywords:mobile robot;movement methods;bayesian framework;active perception;passive perception;geomorphic recognition;vibration data;outdoor geomorphology 收稿日期：2020-07-12. 移动机器人行驶在复杂的环境中，如：火星和 基金项目：国家自然科学基金项目(61773147，U1509203:浙江 省自然科学基金项目(LR17F030005). 月球表面、火灾和战争现场等，在这些危险地带 通信作者：刘华平.E-mail:hpliu(@tsinghua.edu.cn 或者是遥远的不适合远程控制的时候，机器人DOI: 10.11992/tis.202007025 室外未知环境下的 AGV 地貌主动探索感知 张威1 ，葛泉波2 ，刘华平3 ，孙富春3 （1. 上海海事大学 物流工程学院，上海 201306; 2. 同济大学 电子与信息工程学院，上海 201804; 3. 清华大学 计算 机科学与技术系 ，北京 100084） 摘 要：智能机器人对复杂地貌环境的识别一直是机器人应用领域研究的前沿问题，移动机器人在不同的地貌 上采取的运动方式并非一成不变，所以选择的运动方式对于迅速准确识别所处地貌的类型至关重要。针对该 问题本文提出了一种基于贝叶斯框架的主动感知探索方法，使移动机器人能够主动探索有兴趣的运动方式并 且感知识别和运动之间的匹配关系，可以优化在地貌识别之中的模糊不确定性；为了进一步验证实验的可靠 性，还使用了被动感知策略来比较和分析不同策略之间的差异。实验结果表明：主动感知方法能够规划出有效 的地貌识别动作序列，能够引导移动机器人主动感知目标地貌，该框架对于室外未知环境下主动感知后的地貌 识别效果优于被动感知。 关键词：移动机器人；运动方式；贝叶斯框架；主动感知；被动感知；地貌识别；振动数据；室外地貌 中图分类号：TP391 文献标志码：A 文章编号：1673−4785(2021)01−0152−10 中文引用格式：张威, 葛泉波, 刘华平, 等. 室外未知环境下的 AGV 地貌主动探索感知 [J]. 智能系统学报, 2021, 16(1): 152–161. 英文引用格式：ZHANG Wei, GE Quanbo, LIU Huaping, et al. AGV active landform exploration and perception in an unknown outdoor environment[J]. CAAI transactions on intelligent systems, 2021, 16(1): 152–161. AGV active landform exploration and perception in an unknown outdoor environment ZHANG Wei1 ，GE Quanbo2 ，LIU Huaping3 ，SUN Fuchun3 (1. Logistics Engineering College, Shanghai Maritime University, Shanghai 201306, China; 2. School of Electronics and Information Engineering, Tongji University, Shanghai 201804, China; 3. Department of Computer Science and Technology, Tsinghua University, Beijing 100084, China) Abstract: The recognition of a complex landform environment by an intelligent robot has been the frontier problem of research in the field of robotics applications. The motion modes adopted by mobile robots differ between landforms, so the selected motion mode is crucial for quickly and accurately identifying the type of landform. To solve this problem, an active perception exploration method is proposed in this paper based on a Bayesian framework. It enables mobile ro￾bots to actively explore interesting motion modes and recognize the matching relationship between landform and move￾ment. It can optimize the fuzzy uncertainty in landform recognition. To further verify the reliability of the experiment, we also use a passive recognition strategy to compare and analyze the differences between different strategies. The ex￾perimental results show that the active perception method can plan effective landform recognition action sequences and guide mobile robots to actively perceive the target landform. The landform recognition effect of active perception is bet￾ter than that of passive perception in an unknown outdoor environment. Keywords: mobile robot; movement methods; bayesian framework; active perception; passive perception; geomorphic recognition; vibration data; outdoor geomorphology 移动机器人行驶在复杂的环境中，如：火星和 月球表面、火灾和战争现场等，在这些危险地带 或者是遥远的不适合远程控制的时候[1-2] ，机器人 收稿日期：2020−07−12. 基金项目：国家自然科学基金项目 (61773147，U1509203)；浙江 省自然科学基金项目 (LR17F030005). 通信作者：刘华平. E-mail：hpliu@tsinghua.edu.cn. 第 16 卷第 1 期 智 能 系 统 学 报 Vol.16 No.1 2021 年 1 月 CAAI Transactions on Intelligent Systems Jan. 2021