【作者】 方兴；

【导师】 杨明；

【作者基本信息】 上海交通大学 ， 控制理论与控制工程， 2009， 硕士

【摘要】 在行车安全和交通效率问题日益凸显的今天,智能车技术正在快速发展。为了达到更好的车辆控制效果,期望使用基于动力学的车体控制方法。但是在车辆动力学模型中,某些参数的确定非常困难或者测量成本昂贵。为此,本文进行了针对智能车的动力学模型参数辨识研究。为了便于智能车动力学模型参数辨识的研究,本文首先搭建了以智能小车实验平台,该智能小车具备良好的转向特性和速度特性,而且在结合了嵌入式微控制器、摄像头和测速编码器后能够实现车辆的自主导航。然后提出了基于激光雷达的智能小车定位方法,该方法把切线角度直方图和XY直方图的优势进行巧妙结合以进行车辆位姿预估,再运用基于模板匹配和鲁棒迭代最近点算法(Iterative Closest Point, ICP)的定位方法对智能小车进行定位,最后使用扩展卡尔曼滤波(Extended Kalman Filter, EKF)进行位姿跟踪。仿真实验和实际实验证明,该定位方法具备较好的定位鲁棒性和精度。此后,基于以上定位结果,进行了智能小车动力学参数辨识的初步研究,在二自由度动力学模型的基础上推导出参数辨识的方法和步骤,最后基于辨识参数的仿真轨迹与实际的定位轨迹对比,验证了该辨识方法的可行性。更多还原

【Abstract】 As the prominence of traffic safety and efficiency, intelligent vehicle technology develops rapidly. To get the better driving, it’s reasonable to control based on vehicle dynamics. Unfortunately, some parameters in dynamic model are difficult to determine or costly to measure. Research on identification of dynamic parameters for intelligent vehicle is stated in this paper.For the convenience of research work, an experiment platform based on smart vehicle is developed. The smart vehicle performs well in steering and speed property. After being integrated with embedded micro-controllers, camera and encoder, it have the ability of autonomous navigation.Then a localization method of the smart vehicle based on laser radar is given. This method combines the advantages of tangent angle histogram and XY histogram to pre-estimate the pose. Later, model registration and robust Iterative Closest Point (ICP) algorithm is designed to locate the smart vehicle. Finally, for compensating laser radar deficiencies, Extended Kalman Filter (EKF) algorithm is used for Laser Radar and Dead Reckoning data fusion so as to guarantee the accuracy of localization. Simulation and real experiments illustrate that the localization method has high robustness and accuracy.Based on the location result, preliminary and basic research work is done about identification of the smart vehicle. The parameter identification method is derived from the 2-DOF dynamic model. Contrast of simulation trace based on identified parameters and real trace based on localization indicates the feasibility of the method.更多还原