CASIA OpenIR  > 毕业生  > 博士学位论文
智能轮椅控制及人机接口技术研究
其他题名Research on Control and Human-Machine Interfaces for Intelligent Wheelchair
鲁涛
学位类型工学博士
导师原魁
2007-06-01
学位授予单位中国科学院研究生院
学位授予地点中国科学院自动化研究所
学位专业控制理论与控制工程
关键词智能轮椅 嵌入式 分享导航 模糊积分 径向基函数网络 Intelligent Wheelchair Embedded Shared Navigation Fuzzy Integral Rbf Network
摘要随着社会老龄化问题的日益突出,为老年人和残疾人提供生活自理能力是整个社会不得不认真面对并努力解决的一个重要问题。智能轮椅作为一种安全、智能的代步工具,受到了各国研究者的重视。本文针对智能轮椅的控制系统及其人机接口技术进行了深入研究,论文的主要内容如下: 第一,针对以PC机为中央控制器的中央集中控制系统信息处理负担重、实时性差的缺点,研制开发了基于DSP的嵌入式智能轮椅控制系统。系统分为主控系统和传感器系统。主控系统负责系统状态的检测和设置,人机接口信息和传感器信息的融合及运动策略的规划;传感器系统则负责传感器信息的获取及预处理。主控系统和传感器系统通过串行通讯的方式通讯。整个系统具有功耗低,实时性高等特点。 第二,研究了智能轮椅避障及在狭窄区域中的导航策略。在避障情况下,主要分析了在直行和后退转弯过程中使用者的行使意图,并根据不同的情况进行相应的轨迹规划和修正。在狭窄区域情况下,针对门型和走廊型两种情况的转弯路径分别进行了几何分析,并通过反馈线性化的方法实现了路径跟踪。通过仿真和实验验证了导航策略的有效性。 第三,研究了基于头部动作的智能轮椅运动控制。采用自行研制的项圈作为人机接口,通过检测头部运动引起的项圈内弯曲传感器形变达到判断头部运动的目的。针对弯曲传感器特性的非线性特点,采用LMBP网络对传感器进行特性逼近。对于采集到的弯曲传感器信息,则通过模糊积分的信息融合方法对其进行信息融合以判断头部的动作,并进而转化为驱动指令。实验结果证明了该方式的可行性。 第四,研究了基于手势的智能轮椅运动控制。以数据手套作为人机接口,通过定义9种简单的手势来控制轮椅在各个方向上的运动。利用径向基函数网络学习能力强、逼近能力强的特点来进行静态手势识别。针对径向基函数网络中基函数中心点难以确定的特点,提出采用进化规划的方法来自主确定。实验结果证明经过训练的径向基网络对手势的识别效果较好,同时利用手势达到了实时控制轮椅运动的目的。 最后,总结了全文的研究成果,并进一步展望了下一步开展的工作。
其他摘要Firstly, considering the central control system based on PC has heavy information-processing tasks and poor real-time characteristics, a DSP-based control system of intelligent wheelchair is developed. The system includes two subsystems: main subsystem and sensor subsystem. Main subsystem takes charge of state checking and set of system, information fusion from HMI (Human-Machine Interface) and sensors, and motion strategy planning. On the other hand, sensor subsystem takes charge of senor information acquisition and pre-processing. The interconnection of two subsystems is through SCI (Serial Communication Interface). The whole system is low-power-cost and real-time. Secondly, strategies of obstacle avoidance and navigation in narrow spaces for intelligent wheelchair are presented. In case of obstacle avoidance, the dissertation mainly analyzes the user’s drive intention in driving forward and turning backward, and then presents corresponding track planning or track correction according to different conditions. In the narrow spaces, the running conditions of the intelligent wheelchair are divided into two types: the corridor type and the pass-door type, and the path tracking strategy is realized by feedback linearization method.Experiment shows the proposed navigation strategy works quite well. Thirdly, a self-made chaplet is introduced as the head-oriented interface for intelligent wheelchair. The chaplet is mounted flexible sensors which can sense the head movement. A LMBP neural network is used to approach the sensor’s characteristic by taking account of the unregular nonlinear relationship between resistance values and bending angle of flexible sensor. Furthermore, Fuzzy integral method is proposed to fuse the flexible sensor information from the chaplet for the recognization of the head movement, and then the result of recognition is translated into wheelchair’s corresponding motion. The effectiveness of method is verified by experiment results. Fourthly, a data glove as another interface for intelligent wheelchair is detailed. According to motion characteristics of wheelchair, we define nine gestures, each of which corresponds to one motion of wheelchair. A trained RBF network (Radial Basis Function Network) is used to recognize the different gestures.Considering it is hard to confirm the centres of basis functions, EP (Evolution Programming) method is employed to find the centres automatically. In the end, the gestures are translated into drive orders. The experimental result displays that the trained RBF network has a high recognition rate for gestures, and it is effective to control the wheelchair’s motion by data glove in real-time. Finally, the achievements of the research are summarized and the further work is addressed.
馆藏号XWLW1126
其他标识符200418014628002
语种中文
文献类型学位论文
条目标识符http://ir.ia.ac.cn/handle/173211/5993
专题毕业生_博士学位论文
推荐引用方式
GB/T 7714
鲁涛. 智能轮椅控制及人机接口技术研究[D]. 中国科学院自动化研究所. 中国科学院研究生院,2007.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
CASIA_20041801462800(10960KB) 暂不开放CC BY-NC-SA请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[鲁涛]的文章
百度学术
百度学术中相似的文章
[鲁涛]的文章
必应学术
必应学术中相似的文章
[鲁涛]的文章
相关权益政策
暂无数据
收藏/分享
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。