【作者】 王鸿鹏；

【导师】 刘景泰；

【作者基本信息】 南开大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 随着智能移动机器人技术的快速发展,轮式移动机器人在科考、排险、救援、作战、智能交通等领域的应用需求日益增长,已经在各种场合辅助或替代人类以完成特定的任务。轮式移动机器人的运动多是在室外自然环境下进行的。由于环境的复杂性,机器人的自主运动或者遥操作都十分困难,因此迫切需要提升移动机器人在不确定环境下自主定位和运动控制的能力。本论文的研究内容来源于国家自然科学基金项目“基于机器视觉和惯性测量的轮式滑动转向移动机器人定位导航与遥感知”,是其中理论研究的重要组成部分。针对复杂环境下轮式移动机器人的作业要求,对移动机器人的运动机理、自主定位、运动控制等一系列重点问题和关键技术进行了研究。本论文主要由两个部分组成,第一部分针对一般意义下的复杂环境,如起伏地面和狭窄通道等展开研究。平台选用轮式滑动转向移动机器人(WheeledSkid_Steer MobileRobots,WSMR),WSMR的机构简单、性能可靠,有良好的驱动性和通过性。通过对WSMR的运动机理、定位方法、控制方法进行研究,可以提高自主移动机器人在这类环境下的运动性能。第二部分针对山区道路这种典型的三维室外复杂环境,研究数字高程模型辅助的自主定位方法,来提升在无GPS的情况下移动机器人自主导航的能力。轮式滑动转向移动机器人是一种全固定式四轮驱动机构的移动机器人,是非线性、时变、多变量、强耦合的系统,不能用简单的运动学和动力学模型来表示。本文在对WSMR的运动机理分析时,建立了运动打滑模型,在进行运动学分析时充分考虑到了打滑对机器人运动学的影响。在进行动力学分析时讨论了打滑因素对移动机器人运动平稳性的影响,提出并推导出了机器人运动的平稳性判据,得到了机器人保持平稳运动的边界条件。用机器人装配的多个异类传感器采集信息并进行融合来定位是移动机器人常用的自定位方法。尽管全球定位系统(Global Positioning System,GPS)在移动机器人导航定位方面有着广泛的应用,但是不依赖于GPS的移动机器人定位技术仍旧非常关键。本文中利用电机码盘、惯性测量单元、姿态参考系统等异类传感器采集数据,基于扩展卡尔曼滤波方法,将多传感器信息融合定位,在定位过程中引入运动打滑模型和传感器测量误差模型进行实时修正。实验结果表明该组合定位方法可以有效地提高WSMR在起伏地面运动的定位精度。WSMR非完整约束系统的模型不确定性给运动控制提出了挑战。本文采用自适应控制方法,在实时估计车轮与地面间摩擦力的同时,控制移动机器人渐进跟踪期望轨迹,并利用Lyapunov稳定性理论证明了控制器设计的稳定性。本文还提出了应用特技运动控制方法,来实现移动机器人高速精确转向的效果。本文中提出了一种利用数字高程模型数据辅助的轮式移动机器人运动定位方法。对于山区道路环境下的移动机器人定位,当无GPS信号时,航迹推测方法很难奏效。本文从存储于移动机器人本体上的山体数字高程模型数据与遥感影像数据中得到三维路网信息,通过对机器人运动轨迹和路网的几何特征进行匹配来实现移动机器人在山区道路环境下运动的在线全局定位,为移动机器人在山区环境下的作业提供理论支持。为了开展上述研究,本文构建了TamuBot和NKRover-1两套轮式滑动转向移动机器人实验平台,分别对该机器人在三维起伏地面的自定位和平面狭窄跑道中的特技运动控制进行了实验研究。为了测试机器人的定位精度,构建了基于全局视觉的轨迹跟踪系统。本文还进行了机器人山区道路环境下的自主定位实验,对地理信息数据和机器人轨迹进行实时匹配来实现在线的全局定位。更多还原

【Abstract】 With the rapid development of Intelligent Mobile Robot Technology, the wheeled mobile robots are handling many tasks instead of person, for satisfying the demand in scientific expedition, Removal of Dangerous, relief, Battle, and Intelligent traffic areas.Mostly of working range for wheeled mobile robots are outdoor environments. The localization and tele-operation become very hard because of the complex environ-ments. Therefore, It’s very urgent and important for mobile robots to improve the ability of self-localization and motion control in uncertain environments.All the works in the dissertation are sponsored by project fund from National Natural Science Foundation of China. In this paper, the movement principles, method of self localization and motion control were studied.The dissertation is mainly consists of two components. In the first part, researches are focused on the generally complex environments. such as uneven terrain or narrow spaces. Wheeled Skid-Steer Mobile Robots (WSMR) can fit situations in these environments well. WSMRs have simply mechanism, robust reliability,good driving and passing ability. the movement performance of autonomous mobile robots can be improved based on the research of WSMR’s theorem. In the second part, facing at mountain roads which are the classic outdoor environments, a Digital Elevation Model aided self-location method is presented. the self-navigation ability for mobile robot without GPS signals can be improved using this method.Wheeled Skid-Steer Mobile Robots are driven by four fixed wheels. They are nonlinear, time-varying, Multi-variable and strong-coupling systems. The model of WSMR by simple kinematics and dynamics models can not be expressed easily. In this dissertation, a slippage model was built for WSMR robots,the effects of slippage are considered fully in analyzing the kinematics features. Effects are also discussed for stationarity in dynamic analysis caused by slippage factors. the paper present and derive the stationarity criterion, and obtained the boundary conditions for robots to keep stationarity when moving and turning in high speed.Estimation poses by fusing information from multi-sensors mounted on the robots are general method for mobile robots self-localization. Although Global Positioning System (GPS) are used widely in mobile robots outdoor localization, the technology of localization without GPS are still very crucial. this paper compose information from Encoders, Digital Compasses, Inertial Measurement Units and Attitude and Heading Reference Systems to localization using Extended Kalman Filter Based method. Slippage model and measurement error models are introduced into EKF to correcting estimation errors in real-time. The experimental results imply that this combined method can improve precision of localization on uneven terrain effectively.Challenges to Motion Control are posed Owing to the uncertainty of WSMR model. This paper design an adaptive control method in this dissertation to simultaneously estimate the wheel/ground contact friction information and control the robot to follow a desired trajectory. A Lyapunov based convergence analysis of controller and the estimation of the friction model parameter is presented. Paper also discuss how to control a WSMR robot to turning accurately in high speed by stunt movement in this section.The dissertation present a Digital Elevation Model aided self-location method in this dissertation. when robots are running on mountain roads. The regular method for self-localization works hardly without GPS signals. 3D network of roads are obtained from DEM models and remote sensing images stored in the robot memory. The global position can be localized in real-time by matching the geometry curvature features of movement trajectories and road network.In order to support above researches, this paper build two WSMR platforms, named TamuBot and NKRover,respectively. The self-localization experiments on terrain surface are implemented based on TamuBot Robots, and the stunt control experiments are implemented based on NKRover robot. A Trajectory Tracking system based on global vision was built for testing the precision of localization. this dissertation also implement some experiment to validate localization method based on geometry feature matching for mobile robot in mountain roads environments.更多还原