基于变分稀疏高斯过程的多机器人协同感知与围捕

曹凯; 陈阳泉; 魏云博; 刘志; 陈超波; 高嵩

doi:10.16383/j.aas.c240406

基于变分稀疏高斯过程的多机器人协同感知与围捕

doi: 10.16383/j.aas.c240406 cstr: 32138.14.j.aas.c240406

曹凯^{1, 2,},
陈阳泉^{1, 2,},
魏云博^1,,
刘志^1,,
陈超波^1,,
高嵩^1,

1.
西安工业大学电子信息工学院西安 710021
2.
加州大学默塞德分校工程学院MESA Lab 默塞德 95343

基金项目: 国家重点研发计划项目(2022YFE0123400), 陕西省科技厅重点项目(2023-ZDLNY-61, 2022QFY01-16), 国家自然科学基金(62303368), 教育部重点实验室基金(202312-IFTKFKT-007)资助

详细信息

作者简介:
曹凯：西安工业大学电子信息工程学院副教授, 主要研究方向为自主系统与智能控制, 机器人集群, 机器人协同控制, 源定位. E-mail: caokai@xatu.edu.cn

陈阳泉：美国加州大学默塞德分校教授, 主要研究方向为机电一体化, 分数阶系统, 智能控制, 信息物理系统, 无人机. E-mail: ychen53@ucmerced.edu

魏云博：西安工业大学电子信息工程学院硕士研究生, 主要研究方向为多智能体集群控制. E-mail: weiyunbo@st.xatu.edu.cn

刘志：西安工业大学电子信息工程学院硕士研究生, 主要研究方向为多机器人编队控制. E-mail: liuzhi@st.xatu.edu.cn

陈超波：西安工业大学电子信息工程学院教授, 主要研究方向为智能控制, 分数阶系统, 故障诊断与容错控制. E-mail: chenchaobo@xatu.edu.cn

高嵩：西安工业大学电子信息工程学院教授, 主要研究方向为自主智能与无人系统, 目标探测与识别, 智能巡检系统. 本文通信作者. E-mail: gaos@xatu.edu.cn

计量
- 文章访问数: 67
- HTML全文浏览量: 36
- 被引次数: 0
出版历程
- 收稿日期: 2024-06-28
- 录用日期: 2024-11-21
- 网络出版日期: 2024-12-12

Multi-robot Collaborative Perception and Capture based on Variational Sparse Gaussian Process

CAO Kai^{1, 2
,},
CHEN Yang-Quan^{1, 2
,},
WEI Yun-Bo^1
,,
LIU Zhi^1
,,
CHEN Chao-Bo^1
,,
GAO Song^1
,

1.
School of Electronic Information Engineering, Xi'an Technological University, Xi'an, 710021
2.
The MESA Lab, School of Engineering, University of California, Merced 95343

Funds: Supported by National Key Research and Development Program of China (2022YFE0123400), Shaanxi Provincial Department of Science and Technology Key Project (2023-ZDLNY-61, 2022QFY01-16), National Natural Science Foundation of China (62303368), and Ministry of Education Key Laboratory Open Fund (202312-IFTKFKT-007)

More Information

Author Bio:
Cao Kai　Associate Professor at the School of Electronic Information Engineering, Xi'an Technological University. His research interest covers autonomous systems and intelligent control, swarm robots, multi-robot collaborative control, and source localization

Chen Yang-Quan　Professor at the University of California, Merced, USA. His research interest covers mechatronics, fractional order systems, intelligent control, cyber-physical systems, UAV

Wei Yun-Bo　Master student at the School of Electronic Information Engineering, Xi'an Technological University. His research interest is intelligent agent swarm control

Liu Zhi　Master student at the School of Electronic Information Engineering, Xi'an Technological University. His research interest is multi-robot formation control

Chen Chao-Bo　Professor at the School of Electronic Information Engineering, Xi'an Technological University. His research interest covers intelligent control, fractional-order systems, fault diagnosis and fault-tolerant control

Gao Song　Professor at the School of Electronic Information Engineering, Xi'an Technological University. His research interest covers autonomous intelligent unmanned system, target detection and recognition. Corresponding author of this paper

摘要

摘要: 针对未知环境下的多机器人环境感知和围捕问题, 提出了一种基于变分稀疏高斯过程回归的分布式感知与围捕算法. 考虑到传统高斯过程回归不适合处理大量数据的问题, 在这项工作中, 首先考虑障碍物的影响, 以引入分离超平面的质心维诺划分算法为机器人动态规划任务区域; 其次, 利用多机器人在任务区域中的移动探索获取环境信息, 并通过变分自由能方法来近似模型的后验分布, 完成对未知环境的感知; 最后, 基于粒子群算法为围捕机器人动态分配围捕点, 实现多机器人的全方位均匀围捕. 通过仿真实验证明, 该算法能够适用于单源、多源以及动态源的围捕, 且能够在保证多机器人编队安全性的同时, 实现较高的迭代速度, 最终成功实现均匀围捕.
- 多机器人 /
- 质心维诺划分 /
- 变分稀疏高斯过程回归 /
- 围捕 /
- 协同感知
Abstract: To solve the problem of multi-robot environmental perception and capture in unknown environments. Proposing a distributed perception and capture algorithm based on variational sparse Gaussian process regression. Considering the issue that traditional Gaussian process regression is not suitable for handling large amounts of data, in this paper, the influence of obstacles is first taken into account to introduce the centroidal Voronoi partitioning algorithm with separating hyperplanes for dynamic planning of the task area for robots; secondly, the movement exploration of multiple robots in the task area is used to obtain environmental information, and the variational free energy method is used to approximate the posterior distribution of the model to complete the perception of the unknown environment; finally, the PSO algorithm is used to dynamically assign capture points for the capturing robots, achieving all-round uniform capture of multiple robots. Simulation experiments prove that this algorithm is applicable to single-source, multi-source and dynamic source environments, and can achieve high iteration speed while ensuring the safety of the multi-robot formation, ultimately successfully achieving uniform capture.
- Multi-robot /
- centroid Voronoi partitioning /
- variational sparse gaussian process regression /
- capture /
- collaborative perception

HTML全文

图 1 总体框架概述

Fig. 1 Overview of the overall framework

下载: 全尺寸图片幻灯片

图 2 避碰维诺单元

Fig. 2 Collision avoidance Voronoi unit

下载: 全尺寸图片幻灯片

图 3 对污染源的均匀围捕策略

Fig. 3 Uniform capture strategy for a source

下载: 全尺寸图片幻灯片

图 4 高斯分布的污染源仿真

Fig. 4 Gaussian distribution of pollution sources

下载: 全尺寸图片幻灯片

图 5 五个机器人使用VS-GPR策略对单污染源进行围捕的过程

Fig. 5 Five robots capture single pollution source by VS-GPR

下载: 全尺寸图片幻灯片

图 6 五个机器人使用ODMV策略对单污染源进行围捕的过程

Fig. 6 Five robots capture single pollution source by ODMV

下载: 全尺寸图片幻灯片

图 7 六个机器人使用VS-GPR策略对两个污染源进行围捕

Fig. 7 Six robots capture two sources by VS-GPR

下载: 全尺寸图片幻灯片

图 8 六个机器人使用VS-GPR策略对四个污染源进行围捕

Fig. 8 Six robots capture four sources by VS-GPR

下载: 全尺寸图片幻灯片

图 9 多机器人围捕动态污染源

Fig. 9 Multi-robot capture dynamic source

下载: 全尺寸图片幻灯片

图 10 室内实验环境

Fig. 10 Indoor laboratory environment

下载: 全尺寸图片幻灯片

图 11 多无人机控制系统结构框图

Fig. 11 Block diagram of a multi-UAV control system

下载: 全尺寸图片幻灯片

图 12 两种不同的实验场景

Fig. 12 Two different experiments

下载: 全尺寸图片幻灯片

图 13 多无人机围捕单目标源实验

Fig. 13 Multi-UAV capture single-source experiment

下载: 全尺寸图片幻灯片

图 14 多无人机围捕多目标源实验

Fig. 14 Multi-UAV capture multi-source experiment

下载: 全尺寸图片幻灯片

表 1 分类指标及相关工作

Table 1 The criteria classification and related work

分类指标	[18]	[19]	[20]	[21]	[22]	本文方法
围捕机器人数量	多机器人	多机器人	多机器人	多机器人	多机器人	多机器人
被围捕目标源数量	单目标源	单目标源	单目标源	单目标源	多目标源	多目标源
2D/3D环境	2D	3D	2D	2D	3D	3D
动态/静态目标源	动态	动态	动态	动态	动态	动态
是否感知场源信息	否	否	否	否	否	是
是否有实验验证	否	否	否	否	是	是

下载: 导出CSV

表 2 仿真参数设置

Table 2 Simulation parameter settings

参数	参数值
任务切换阈值	4.5
障碍物的二维坐标(m)	(5.0, 1.5)(1.2, 3.5)(4.0, 5.0)
障碍物尺寸	0.5m$\times$0.5m$\times$3m
UAV最大速度	0.3m/s
机器人和待围捕点最小距离	0.5m

下载: 导出CSV

表 3 单污染源下两种围捕方法对比

Table 3 Comparison of two capture strategies

围捕方法	ODMV围捕	VS-GPR围捕
运行时间(s)	197.62	102.67
迭代次数	180	70

下载: 导出CSV

表 4 多机器人围捕多污染源用时对比

Table 4 Comparison of time consumption for multi-robot capture multi-source

污染源数量	GPR耗时(s)	VS-GPR耗时(s)
2个	278.63	102.67
3个	319.37	113.41
4个	345.56	118.32
5个	352.17	118.84

下载: 导出CSV

参考文献(22)

[1]	蒲志强, 易建强, 刘振, 丘腾海, 孙金林, 李非墨. 知识和数据协同驱动的群体智能决策方法研究综述. 自动化学报, 2022, 48(3): 627−643 Pu Zhi-Qiang, Yi Jian-Qiang, Liu Zhen, Qiu Teng-Hai, Sun Jin-Lin, Li Fei-Mo. A review of the research on swarm intelligence decision-making methods driven by knowledge and data collaboration. Acta Automatica Sinica, 2022, 48(3): 627−643
[2]	Chung T H, Hollinger G A, Isler V. Search and pursuit-evasion in mobile robotics: A survey. Autonomous robots, 2011, 31(4): 299−316 doi: 10.1007/s10514-011-9241-4
[3]	Zhang J, Lu Y, Wu Y, Wang C, Zang D, Abusorrah A, Zhou M. PSO-based sparse source location in large-scale environments with a uav swarm. IEEE Transactions on Intelligent Transportation Systems, 2023, 24(5): 5249−5258 doi: 10.1109/TITS.2023.3237570
[4]	Karagüzel T A, Turgut A E, Eiben A E, Ferrante E. Collective gradient perception with a flying robot swarm. Swarm Intelligence, 2023, 17(1): 117−146
[5]	Jang H, Park M, Oh H. Improved Socialtaxis for information-theoretic source search using cooperative multiple agents in turbulent environments. Expert Systems with Applications, 2023, 225: Article No.120033 doi: 10.1016/j.eswa.2023.120033
[6]	Chen L, Zhou C, Wang Y, Zong Y, Lu T, Chen C. Autonomous search investigation for radioactive leaked source based on an updated infotaxis method during nuclear emergency rescue. Nuclear Engineering and Design, 2024, 416: Article No.112769 doi: 10.1016/j.nucengdes.2023.112769
[7]	Ma H, Zhang T, Wu Y, Calmon F P, Li N. Gaussian max-value entropy search for multi-agent Bayesian optimization. In: Proceedings of the International Conference on Intelligent Robots and Systems (IROS). Detroit MI, USA: IEEE, 2023. 10028−10035
[8]	Xu Y, Zheng R, Zhang S, Liu M, Huang S. CARE: Confidence-rich autonomous robot exploration using bayesian kernel inference and optimization. IEEE Robotics and Automation Letters, 2023, 8(10): 6755−6762 doi: 10.1109/LRA.2023.3313054
[9]	Jabeen M, Meng Q H, Jing T, Hou H R. Robot odor source localization in indoor environments based on gradient adaptive extremum seeking search. Building and Environment, 2023, 229: Article No.109983 doi: 10.1016/j.buildenv.2023.109983
[10]	Li H, Yuan J, Yuan H. An active olfaction approach using deep reinforcement learning for indoor attenuation odor source localization. IEEE Sensors Journal, 2024, 24(9): 14561−14572 doi: 10.1109/JSEN.2024.3373610
[11]	Latif E, Parasuraman R. Communication-efficient multi-robot exploration using coverage-biased distributed Q-learning. IEEE Robotics and Automation Letters, 2024, 9(3): 2622−2629 doi: 10.1109/LRA.2024.3358095
[12]	Liao Y, Chen A, Cai H, Lu J, Zhang W, Lin B, Tong Y. An experimental study on locating time-varying pollutant sources in a dynamic indoor environment with mechanical ventilation based on multi-robot cooperation. Journal of Building Engineering, 2023, 67: Article No.105962 doi: 10.1016/j.jobe.2023.105962
[13]	Yu L, Huo S, Wang Z, Li K. Hybrid attention-oriented experience replay for deep reinforcement learning and its application to a multi-robot cooperative hunting problem. Neurocomputing, 2023, 523: 44−57 doi: 10.1016/j.neucom.2022.12.020
[14]	Qu X, Gan W, Song D, Zhou L. Pursuit-evasion game strategy of USV based on deep reinforcement learning in complex multi-obstacle environment. Ocean Engineering, 2023, 273: Article No.114016 doi: 10.1016/j.oceaneng.2023.114016
[15]	Zhao Z, Hu Q, Feng H, et al. A cooperative hunting method for multi-AUV swarm in underwater weak information environment with obstacles. Journal of Marine Science and Engineering, 2022, 10(9): 1266 doi: 10.3390/jmse10091266
[16]	Zhou M, Wang Z, Wang J, Cao Z. Multi-robot collaborative hunting in cluttered environments with obstacle-avoiding voronoi cells. IEEE/CAA Journal of Automatica Sinica, 2024, 11(7): 1643−1655 doi: 10.1109/JAS.2023.124041
[17]	Wang X, Xi L, Ding Y, Chen B M. Distributed encirclement and capture of multiple pursuers with collision avoidance. IEEE Transactions on Industrial Electronics, 2023, 71(7): 7520−7530
[18]	Li B, Wang J M, Song C, Yang Z P, Wan K F, Zhang Q F. Multi-UAV roundup strategy method based on deep reinforcement learning CEL-MADDPG algorithm. Expert Systems with Applications, 2024, 245: 123018 doi: 10.1016/j.eswa.2023.123018
[19]	Zhao Z Y, Zhang Y Z, Feng X L, Jiang C, Su W B, Hu Q. A dynamic velocity potential field method for multi-AUV cooperative hunting tasks. Ocean Engineering, 2024, 295: 116813 doi: 10.1016/j.oceaneng.2024.116813
[20]	Fu X W, Zhang Y X, Zhu J D, Wang Q L. Bioinspired cooperative control method of a pursuer group vs. a faster evader in a limited area. Applied Intelligence, 2023, 53(6): 6736−6752 doi: 10.1007/s10489-022-03892-8
[21]	Tian B, Li P, Lu H, Zong Q, He L. Distributed pursuit of an evader with collision and obstacle avoidance. IEEE Transactions on Cybernetics, 2021, 52(12): 13512−13520
[22]	Cao K, Chen Y Q, Gao S, Yan K, Zhang J H, An D. Omni-directional capture for multi-drone based on 3D-Voronoi tessellation. Drones, 2023, 7(7): 458 doi: 10.3390/drones7070458