【作者】 徐凯；

【导师】 陈小平；

【作者基本信息】 中国科学技术大学 ， 计算机应用技术， 2008， 博士

【摘要】 由于结构上的优势,自主步行机器人相对于履带式机器人和轮式机器人具有更广泛的适用范围。自主步行机器人往往工作在动态不确定环境中,这对机器人的运动能力、对环境的适应性和灵活性、对自身运动状态的实时辨识能力提出了很大的挑战。过去对步行机器人的运动设计、行走状态稳定性等的研究取得了一系列的成果,但随着机器人技术的发展所带来的机器人平台性能的增强以及机器人所面临任务的复杂化,对自主步行机器人的行走设计、运动状态分析及标定能力提出了更高的要求。为了提升步行机器人的行走速度,以及在高速行走时对自身运动的快速理解识别能力,从而提高机器人在未知环境中的工作能力,本文以AIBO四足机器人为研究平台,从以下三个方面对自主步行机器人的运动控制进行了研究:1.步行机器人的步态轨迹生成及优化以前的基于步态规划的步行机器人行走设计,在完成机器人腿部建模及相应的逆运动学解算后,通常是采用如矩形、梯形等固定轨迹设计机器人腿部的运动,缺乏对机器人行走的最优步态轨迹的研究,难以保证其行走设计在各种不同的条件和环境下都能达到较好的结果。本文在步行机器人步态设计基础上,提出了采用曲线拟合的方法在机器人肢体运动空间内生成步态轨迹曲线,通过引入遗传算法,让机器人能够自主的搜索良好的步态轨迹,增强了机器人的运动能力。2.步行机器人的行走状态分析模型行走速度的提高对步行机器人行走的稳定性提出了更高的要求,已有的研究主要关注机器人在行走是否发生翻转,缺乏对碰撞定量描述、行走稳定程度等方面的考虑。针对这一问题,结合步行机器人行走的动力学特性,本文对机器人的加速度传感器信息进行离散傅立叶变换,建立了行走相关特征值的概率模型,通过使用马氏距离作为判定标准,对步行机器人的行走稳定性给出定量描述。AIBO四足步行机器人平台上的实验证明,本文提出的模型能够实时反应机器人的行走特性,帮助机器人在行走状态受环境影响发生改变时,根据行走特征及时调整运动,保证其稳定性。3.自标定的步行机器人行走状态模型相对轮式机器人和履带式机器人,步行机器人的行走往往更为复杂,为了准确标定其运动状态,特别是运动频繁改变时,必须提高机器人运动中的在线自标定能力,以增强其对自身状态的快速理解识别能力。因此,以提高机器人对自身行走状态模型的标定能力为出发点,本文结合文献[79]提出的运动模型和本文建立的基于加速度传感器信息的行走状态传感器模型,通过运动模型和传感器模型间的互标定,使步行机器人能够在正常行走下在线完成对行走状态的自标定。本文以AIBO四足机器人作为实验平台,分别对上述三方面的研究进行了实验并取得了良好的实验结果。实验结果表明,本文所提出的步态轨迹生成方法能有效的提高步行机器人的行走速度,基于加速度传感器信息的行走状态模型可以定量描述步行机器人运动的稳定性,自标定的行走状态模型能够让步行机器人在行走时在线标定自身行走状态。通过上述三方面的工作,本文在自主步行机器人的行走设计、优化,行走状态的辨识,以及行走状态模型的自标定进行了研究,这些研究结果完善了步行机器人的运动控制,增强了机器人对自身运动状态的理解能力以及对未知环境的适应能力。更多还原

【Abstract】 Because of the advantage of structure, the work area of autonomous legged robot is broader than that of crawler robot and wheeled robot. Autonomous legged robot often works in dynamic uncertain environment; it’s a great challenge for legged robot in locomotion, flexibility and real-time walking status recognition. A lot of research results have been achieved in locomotion design and walking stability. But, with the development of robot technology and the enhanced performance of robot platform, the task is also more complicated than before, and the higher demand is put forward in walking design, walking status analysis and calibration for autonomous legged robot. In order to improve the walking speed of legged robot and increase the capability of legged robot to understand its own status in high-speed movement to enhance robot work ability in unknown environment, with the research platform, quadrupedal robot AIBO, this paper studies the locomotion control of autonomous legged robot in the following three aspects:1. The gait locus generation and optimization of legged robotIn the previous walking design of legged robot which is based on gait design, the rectangle locus and trapezoid locus are often used after the leg model building and inverse kinematic, the lack of research of optimization gait locus makes it difficult for the walking design to work well in different environments and conditions. Based on the gait design, this paper presents a curve fitting planning method to generate the gait locus in the work space of the limb of legged robot while incorporating genetic algorithms, it makes legged robot search gait locus automatically to enhance the robot movement ability.2. The walking state analysis model of legged robotWith the development of legged robot walking speed, the higher demand of walking stability has been put forward. Many research results concentrate on whether turnover is occurred, lack of ideas about the quantitative description of collision and walking stability. Bearing this problem in mind, this paper presents a possibility walking state analysis model for legged robot which is based on the feedback of acceleration sensor. The feedback of acceleration sensor is processed by Discrete Fourier Transform while incorporating the dynamic analysis of the walking of legged robot firstly, then, according to be estimated by Mahalanobis Distance, the result can express the legged robot walking status quantitatively. Using quadrupled robots as a platform for evaluation, this model is shown to be able to descript the walking of legged robot and express the walking status in real time, and it could be helpful for legged robot to adjust its locomotion with the change of environment to ensure the walking stability of legged robot.3. A self-calibration walking state model of legged robotThe walking of legged robot is more complicated than crawler robot and wheeled robot, the walking state calibration for legged robot is very important, especially when motion is changed rapidly, so the on-line walking state model self-calibration capability of legged robot must be enhanced to increase the capability of understanding its own status in high-speed movement. With the idea of improving the calibration capability of legged robot, by combining the motion model of paper[79] and the walking state model which is based on acceleration sensor, this paper presents a self-calibration model for the walking state of legged robot. Utilizing the inter-calibration between motion model and sensor model, legged robot can on-line self-calibrate its own walking state.Using quadrupled robots as a platform for evaluation, this paper makes the experimental study in the three research aspects and achieves good results. The experimental results show that the gait locus generation method can improve the walking speed of legged robot effectively, the walking state model which is based the feedback of acceleration sensor can descript the walking stability of legged robot quantitatively, and the self-calibration walking model make legged robot be able to calibrate its own walking state online.Through the three aspects, this paper studies in the walking design, optimization, identification of walking state and self-calibration of walking state model of autonomous legged robot. The research results improve the walking design of legged robot, enhance the performance of legged robot with the capability of understanding its own walking status and make legged robot be able to adapt to different environments.更多还原