【作者】 柳玉甜；

【导师】 蒋静坪；

【作者基本信息】 浙江大学 ， 控制理论与控制工程， 2007， 博士

【摘要】 随着移动机器人技术的发展,移动机器人被应用于许多场合,如:未知领域的探索、搜救行动、军事等。在移动机器人的研究过程中,大部分研究都是建立在系统能够以理想状态工作的基础上。但是,经验表明,移动机器人在真实环境中工作时,发生故障的现象还是很普遍的。尤其是在复杂的未知环境中,移动机器人机械部件和控制系统极易出现故障,同时人类也无法对其进行直接干预(或人类干预的代价太高),如果故障没有检测出来或者没有得到及时的处理,移动机器人将会以一种不可预测和危险的方式工作。这不仅会缩短移动机器人的使用寿命,严重时会使移动机器人不能进行正常活动,甚至可能会导致灾难性的后果。因此,移动机器人故障诊断技术的研究具有重要的理论意义和实际应用价值。本文针对当前移动机器人的故障诊断技术领域的研究现状,结合相关学科理论取得的新成果,研究了若干未知环境中移动机器人的故障诊断方法。本文的主要工作和取得的研究成果如下:1.概述了国内外故障诊断技术的研究现状。对基于数学模型的故障诊断方法、基于信号处理的故障诊断方法和基于知识的故障诊断方法进行了分析讨论。2.详细讨论了未知环境中移动机器人的故障诊断问题。对移动机器人的系统组成、故障分类、各子系统的故障诊断方法进行了研究。并对移动机器人的运动单元建立了运动学和动力学模型。3.提出了一种基于多模型估计方法的神经网络故障诊断技术。建立了移动机器人运动单元的故障模型。提出的故障诊断方法利用小脑模型关节控制器(CMAC)神经网络的分类逼近能力,建立故障征兆和故障类型之间的精确映射,以实现故障诊断任务。仿真实验证明了该故障诊断技术在移动机器人故障诊断上的有效性和可行性。4.提出了一种基于粒子滤波器的算法,用于解决移动机器人系统(包括离散故障状态和连续状态)故障诊断和预测问题。基于粒子滤波器的故障诊断和预测算法通过一组带权值的粒子来估计系统的状态,以此计算故障状态的分布情况和故障发生的概率,从而判断是否发生故障以及所发生的故障类型。仿真结果表明,所提出的故障诊断和预测方法能有效地诊断和预测移动机器人的故障模式。5.提出了一种处理不同运动状态下移动机器人故障的诊断方法。该故障诊断方法将移动机器人的运动状态进行分类。在不同的运动状态下,考虑不同的故障模式,并用卡尔曼滤波器组处理不同运动状态下不同模式的故障发生的概率,根据故障发生的概率值判断故障模式发生的可能性。与其他故障诊断方法相比,该故障诊断方法有效改善了误诊和漏诊现象。仿真结果表明在移动机器人故障诊断上的有效性和可行性。最后,对本文的研究工作进行了总结,并对移动机器人故障诊断的未来研究方向作了展望。更多还原

【Abstract】 With the remarkable progress in robotics, mobile robots can be used in many applications including exploration in unknown area, search and rescue, reconnaissance, security, military, rehabilitation, pick up and delivery, and cleaning. Most of the research work on mobile robots has been done in which the system in ideal. Experience with physical mobile robots has shown that robot failure is very common. Especially in complex and unknown environments, mechanism components and control systems of robots are possibility to become faulty, where human intervention is expensive, slow, unreliable, or impossible. If the faults can’t be addressed, or processed in time, mobile robots will operate in an unpredictable and dangerous way. Faults will decrease the performance of the robots, and make the robots can’t operate in normal manner, even will result in catastrophic failures. It is therefore important for robots to monitor their state so that anomalous situations may be detected in a timely manner.According to the current international research situation of fault diagnosis technique for mobile robot, combined with new results obtained in relative disciplines, some kinds of fault diagnosis schemes for mobile robot in unknown environment are proposed in this thesis. The major work and the result of research are represented as follows:1. In this thesis, the current international research situation of fault diagnosis technique is summarized. The mathematics model, signal processing and knowledge based fault diagnosis method are studied.2. The fault diagnosis of mobile robot in unknown environment is studied. The subsystems of mobile robot are introduced. The fault modes of each subsystem are given. The fault diagnosis methods of each subsystem are summarized. Then, the kinematics model and dynamic model of mobile robot are constructed.3. Based on multiple model estimation and Cerebellar Model Articulation Controller (CMAC) neural network, a new fault diagnosis method for mobile robots is studied. Firstly, the fault models of mobile robots motion system are constructed. Then, using the sort approximation ability of the neural network, an exact mapping from space of fault symptom to space of fault modes is established. Finally, the proposed fault diagnosis method apply the CMAC neural network to decide which fault has occurred, This method has been implemented on a mobile robot and the simulation results show the effectiveness of the method. 4. An improved particle filter based approach is proposed to diagnosis and prediction the fault modes of mobile robots. Particle filter is a algorithm that uses swarms of weighted particles in state space to approximate the probability density function of the state. According to the probability distribution of the state, the probability of fault diagnosis and fault prediction could be obtained. The proposed approach has been implemented on a mobile robot and the simulation results show the effectiveness of the method.5. A new method is introduced to detect and diagnose the faults of mobile robots in different movement states. The movement states of mobile robot include static state, rectilinear movement state, and three kinds of turning states. Several modes of faults are discussed in the corresponding movement states. Then a bank of Kalman filters are used to process the mode probability of each fault mode occurring during the movement states. According to the values of mode probability, we can estimate which mode of faults occurred. Compared with other fault detection and diagnosis methods, the method proposed in this paper improves the capability of avoiding the appearance of misdiagnosis and failing to detection and diagnosis. This proposed method has been implemented on a mobile robot and the simulation results show the effectiveness of the method.Finally, the work of this thesis is summarized and the prospective of future research is discussed.更多还原