【作者】 王仲民；

【导师】 岳宏；

【作者基本信息】 河北工业大学 ， 机械制造及其自动化， 2006， 博士

【摘要】 本文针对移动机器人运动规划研究领域中的路径规划问题，并以非完整轮式移动机器人(WMR)作为研究对象，对其轨迹跟踪控制等主要问题进行了深入探索与研究，主要包括以下四部分内容： 静态结构化确定性环境中移动机器人全局路径规划研究。基于神经网络与碰撞罚函数建立了路径与障碍物间的优化模型，并利用模拟退火算法对其进行求解。针对经典模拟退火算法收敛速度慢这一缺陷，首先对该算法进行了改进，提出了一种新型改进模拟退火算法。其次，基于模拟退火算法提出了两种新型混合全局优化算法，有效提高了收敛速度，满足了系统实时性的要求，实现了移动机器人的全局路径规划。最后，应用DFP变尺度与微粒群混合串行算法结构，快速完成了移动机器人全局路径规划任务。 基于Q强化学习与CMAC神经网络的移动机器人局部路径规划研究。移动机器人通过强化学习感知环境状况，选择恰当的行为，并从环境中获得不确定奖赏值作为学习动态系统的最优行为策略。采用CMAC网络实现了Q强化学习算法，在调整CMAC网络权值时，充分考虑已学习数据的可信任度，显著提高了传统CMAC在线学习的速度与准确性，有效地实现了在复杂环境中的局部路径规划，并且效果明显。 基于自适应控制、最优控制以及模糊控制等控制理论与方法，研究非完整轮式移动机器人的轨迹跟踪问题，使移动机器人实现合理乃至最优的运动。(1)基于Lyapunov函数，设计新型轮式移动机器人速度跟踪控制律，实现了对于期望轨迹的准确跟踪。(2)在考虑驱动电机动态模型的前提下，设计出能够使移动机器人系统全局渐近稳定的自适应控制器，有效克服了模型误差以及外部扰动等不确定性因素给系统带来的影响，并进行了相应的轨迹跟踪仿真实验。(3)基于LQR理论设计了移动机器人控制器。针对加权阵Q与R难以确定的问题，从控制效果出发，采用自适应遗传算法对该控制器进行了优化，实现了对于预定轨迹的准确跟踪。(4)基于模糊控制方法实现了移动机器人的鲁棒轨迹跟踪。采用复合形算法对模糊控制器进行了优化，消除了系统的稳态误差，系统不仅具有良好的动特性，而且取得了满意的跟踪效果。 非完整WMR的实验研究。首先详细叙述了自行研制的两轮驱动WMR实验样车及相应的控制系统，然后在该实验平台上进行了WMR的轨迹跟踪实验，结果验证了提出的轨迹跟踪控制策略的正确性与可行性。更多还原

【Abstract】 This paper primarily takes nonholonomic wheeled mobile robot (WMR) as researching object. It does some in-depth researches on path planning and trajectory tracking controlling in the field of motion planning of mobile robot. It mainly includes four parts as follows:Research of mobile robot global path planning in static structural certain environment. Based on neural networks and collision punish function an optimization model is established, and then the model is solved by using simulated annealing algorithm (SAA). It is well known to all that convergence speeds of classical SAA are very slow. So aiming at improving the convergence speeds of SAA, this algorithm is modified at first, then a new modified SAA (MSAA) is put forward. Secondly, two new hybrid optimal algorithm are brought forward subsequently. They satisfy the demands of real time of system all together and realize global path planning of mobile robot, also. Finally, by utilizing DFP variable metric method combined with particle swarm optimization (PSO) algorithm, the tasks of global path planning are accomplished quickly.Research on local path planning of mobile robot based on Q reinforcement learning and CMAC neural networks. The robot senses the states of environment and selects right behavior through reinforcement learning. It is some uncertain prize value acquired from the environments that become the optimum behavior strategy. By using CMAC networks the Q reinforcement learning is realized. And credit assignments of the data, which have being learned, must be fully considered when adjusting the weight of networks. So it obviously improves online learning speed and its accuracy of conventional CMAC. Finally, local path planning in complex environments are completed and the effects are good.Research on trajectory tracking problems of nonholonomic WMR by using control theory and method of adaptive control, optimal control and fuzzy control, etc. It drives the mobile robot to achieve results of perfect so much as optimal controlling at last. (1) A new velocity tracking control strategy is designed based on Lyapunov function. By using this the WMR fulfills to track an expectation trajectory exactly. (2) Including motor dynamics model, an adaptive controller, which can make the system of mobile robot be global asymptotically stable, is presented. It can get rid of influences of modeling errors and external disturbances at the same time. And some simulation experiments are carried out later. (3) An optimal controller is designed based on linear quadratic regulator (LQR). Because it is difficult to determine the weighted matrix Q and R, so更多还原