SIA OpenIR  > 机器人学研究室
基于神经元群的网络遥操作机器人控制研究
Alternative TitleInternet Teleoperation Robot Control based on Neural Group Method
李响1,2
Department机器人学研究室
Thesis Advisor王越超
ClassificationTP242
Keyword网络遥操作机器人 随机时延 基于事件控制 神经元群 多层延迟结构。
Call NumberTP242/L35/2010
Pages89页
Degree Discipline模式识别与智能系统
Degree Name博士
2009-06-30
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract遥操作技术是机器人学中的一个重要研究方向,互联网的出现和快速发展为本地机器人系统和远端操作者之间提供了新的交互媒介。通过网络将操作者智能与机器人局部自主相结合,充分利用操作者与机器人系统间的交互能力,实现远距离、动态非线性环境下的复杂任务操作,极大地拓展了遥操作机器人系统的应用范围,具有广阔的应用前景。但由于网络传输特性产生的随机时延和多通道问题会严重降低遥操作系统的稳定性;环境的复杂性和动态性也会影响系统的操作性能,使得操作者与机器人之间难以实现良好的交互和控制,甚至威胁到机器人自身的安全。针对这一问题,本文在分析了现有的遥操作理论和方法的基础上,采用事件驱动和动态神经元群控制相结合的方法,研究网络遥操作机器人系统的建模和控制问题。通过提高遥操作机器人的局部自主能力,结合学习能力、短期记忆结构和内部表达方式的建立,增强遥操作机器人的环境适应能力。论文的主要研究内容可分为如下三个方面:基于事件和神经元群方法的控制系统整体设计;具有局部自主的神经元群控制器分析与实现;将多层延迟结构引入神经元群模型,给出了动态神经元群控制器设计。具体研究内容概括如下: 1. 论文对遥操作机器人系统发展历程、研究现状,以及在网络遥操作机器人系统研究中所涉及的一些主要方法,存在的问题和已有的解决方案进行了归纳总结和较为全面系统的分析,然后给出本文研究的主要内容和研究所需的一些准备知识。 2. 论述了涵盖三个大脑信息处理特性:选择性、适应性及协同性的神经元群选择理论,详细分析了选择性神经元群网络的结构、功能和算法。建立了基于选择性神经元群网络的遥操作机器人控制模型。在此基础上进行的仿真对比实验,验证了其在学习能力和收敛精度方面,比传统的人工神经网络方法更具优势。 3. 针对网络遥操作机器人系统中存在的随机时延和模型所具有的未知、不确定和非线性特性,造成系统不稳定和操作性能下降等问题,提出采用基于事件和神经元群控制相结合的方式建立系统控制模型。综合考虑两种控制策略的特点,分别设计两端控制器。通过神经元群模型提高本地机器人的局部自主能力,实时跟踪远端的位置和速度信息;结合基于事件控制方法实现切换控制,降低网络随机时延的影响,保证系统稳定性,提高系统操作性能。在此基础上,结合轮式移动足球机器人模型、网络通讯协议和时延数据的分析,进行仿真实验,结果验证了所提控制方法的有效性。 4. 将神经动力学概念引入到网络遥操作系统中,建立一种新的具有分层延迟结构的动态神经元群模型。引入时延单元增加记忆功能,由此导出的模型更适合应用于处理序列数据。动态神经元群模型将事件序列的时变量转化为空间向量表示,进而转化为多维曲线的分类问题。利用这个模型,实现对遥操作机器人系统动态特性参数的辨识和远端操作者意图的分析预测,并进行了全局渐近稳定性和同步状态分析。通过这种方式克服网络随机时延的影响,保证系统的稳定性和操作性能。针对轮式移动足球机器人进行了仿真实验研究。最后,本文分别利用了基于MATLAB和MSVisual C++开发的仿真平台,验证了以上方法的有效性和控制器的性能。本文所作的理论分析工作及仿真实验,验证了基于事件和神经元群方法实现对网络遥操作机器人控制的有效性。
Other AbstractTeleoperation technology is an important subject in the robotics. The emergence and the rapid development of Internet technology provide a new interactive media platform for telerobot and operator. Through Internet, it can combine the operator intelligence and the local-autonomy of telerobot. With the advantage of interactive capability between the operator and the telerobot, it may realize the complex mission under the long-range, dynamic non-linear environment operations, which greatly expand the domain of application. However, network transmission characteristics of the random time-delay, multi-channels would seriously reduce the stability of teleoperation robotic systems. The complexity of dynamic environment and unpredictability will also affect the system performance, making it difficult to achieve the good interaction and control between the operator and the teleoperation system, or even threat the safety of the robot. To solve the above problems, the event-based and dynamics neural group control methods for teleoperation robot system modeling and control are adopted, with analyze the existing theories and methods of teleoperation. By increasing the local autonomy of telerobot, combined with learning ability, short-term memory structure and the establishment of internal expression, to enhance the environment adaptability of telerobot. The main research contents of this paper can be divided into three parts as follows: the event-based and neural group network control method for the remote-local controller design; the analysis and realization of neuron group controller combined with wheeled mobile robot dynamics and kinematics model; the dynamic neuron group controller design with conbination of multi-layer delay structure and neural group model. Then dessertation is organized as follows: 1. We carry out more comprehensive and systematic research and analysis on the history of the telerobot system development, research, and some of the main methods involved. The architecture of Internet-based teleoperation robot control system, the problems and solutions have also been summarized. Then, states the main focuses of this study and prepare the necessary knowledge. 2. Dissert the selective neuron group theory, which covering the three characteristics of brain information processing: selective, adaptive and collaborative. Detailed analysis the structure, functions and algorithms of neuron group network, and establish the control model to the Internet-based telerobot The simulation results verify its convergence in the learning ability and precision, which comparison with the traditional artificial neural network methods. 3. The random time-delay exist in the Intenet-based telerobot systems and the model with the unknown, uncertainty or even non-linear characteristics, making the system instability and decrease the performance issues.We propose to establish the control system model with the event-based and neuronal group method, with considering the characteristics of two control strategies.With the neuron group model, it is able to improve the local-autonomy of telerobot, tracking the real-time position and velocity information. By switching control, it may reduce the effects of random time-delay, ensure the stability of the system and improve system performance. With the combination of wheeled mobile robot model, network communication protocol and the time-delay data analysis, simulation results verify the effectiveness of proposed control method. 4. The concept of neural dynamics is introduced to the Internet-based telerobot control system, and a new hierarchical delayed structure of the dynamic neuron group model is established. The introduction of time-delay unit can increase the memory function, and the derived model is more suitable for handling the sequence data. Time-delay neural group network transfer the time series into a space vector, and then translated into the classification of multi-dimensional curve. With the model, it can realize the dynamic characteristics parameter identification and prediction of teleopration robot system, the intention analysis and prediction of the remote operator. The global asymptotic stability and synchronization analysis is also done. It could overcome the effects of random time delay to ensure the system stability and performance.The simulation experiment carry out based on the wheeled mobile robot system. Finally, to verify the validity of the above methods and the performance of the controller, MATLAB simulation platform and the Internet soccer robot system developed by Visual C + +6.0 are established as experimental platform. In this dessertation, the theoretical analysis and simulation experiments verify the feasibility of the event-based and neuron group control method for Internet-based teleoperation robot system.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/160
Collection机器人学研究室
Affiliation1.中国科学院沈阳自动化研究所
2.中国科学院研究生院
Recommended Citation
GB/T 7714
李响. 基于神经元群的网络遥操作机器人控制研究[D]. 沈阳. 中国科学院沈阳自动化研究所,2009.
Files in This Item:
File Name/Size DocType Version Access License
10001_20031801470306(2246KB) 开放获取--Application Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[李响]'s Articles
Baidu academic
Similar articles in Baidu academic
[李响]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[李响]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.