【作者】 宋金泽；

【导师】 贺汉根；

【作者基本信息】 国防科学技术大学 ， 控制科学与工程， 2009， 博士

【摘要】 自主泊车技术是目前智能车辆技术研究的一个热点问题,因为自主泊车技术的发展有着巨大的市场需求。在与第一汽车集团的合作项目“红旗HQ3自动泊车技术”和国家自然科学基金重点项目“高速公路车辆智能驾驶中的关键科学问题研究”(90820302)的共同支持下,本文重点研究了自主泊车系统的轨迹规划与轨迹跟踪控制问题,具体的研究内容包括:控制系统体系结构、泊车轨迹生成、高精度控制等内容。本文采用理论分析与工程实践相结合的研究方法,实现了高精度自主泊车控制。本文的主要研究成果和创新点如下:1.提出并实现了将分数阶微积分运用于自主泊车控制系统中,解决了自主车辆在低速条件下的高精度侧向控制问题。2.针对“长记忆法”和“短记忆法”在分数阶微分方程数值求解过程中存在的问题,提出一种基于高斯核的分数阶控制系统微积分方程数值计算方法,实践表明该方法具有运算速度快、计算精度高的优点。3.针对分数阶控制器的参数优化问题,提出了一种基于近似策略迭代的分数阶控制器参数自整定方法。在此算法中,针对基函数选择问题,提出了一种基于Bellman残差的基函数自动选择方法,提高了值函数逼近器的逼近精度和泛化能力。4.提出了一种基于人机协同的三层递阶式智能控制系统结构,该结构将人的智能引入到控制系统的每个层次中,能够综合利用人的定性感知、决策、规划能力和机器的定量感知、计算、高精度操作能力,充分地发挥人和机器智能的各自优势,提高了自主泊车系统运行的安全性和环境适应性。5.结合侧位泊车的各种工况,提出了一种平行泊车避碰约束空间生成方法。运用此方法可以方便的确定出泊车过程中的无碰撞区域。提出一种融合微分平坦、样条理论的自主平行泊车轨迹生成方法,可以在上述无碰撞区域生成一条满足运动学、动力学约束的泊车轨迹。实车实验表明,这种方法不仅能够有效地实现避障,而且解决了在泊车过程中的停车原地转方向盘的问题,满足自主泊车的连续性要求。通过大量平行泊车实验,实现了车辆自主平行泊车的控制功能,验证了上述方法的可行性与有效性。本文的研究成果为“红旗HQ3自动泊车技术”课题的完成作出了重要贡献。更多还原

【Abstract】 For its enormous market demand, Autonomous-parking technology is a hot research topic in recent Intelligent Vehicle research area. Under the support of the project‘Hongqi HQ3 Autonomous Parking Technology’in collaboration with Faw Group Corporation and the emphasis project from National Natural Science Foundation of China:‘Research on the key scientific problems in intelligence driving of highway vehicles’(90820302), the key research issues in this thesis are trajectory planning and trajectory tracking&control in the Autonomous Parking system. The specific research issues are: the architecture of the control system, generation of the parking trajectory, high-precision control and etc. Adopting the methodology of combining theoretical analysis and engineering practice, this thesis has achieved the goal of developing a high-precision autonomous parking control system.The main results and innovations of this thesis are as follows:1 Having proposed and implemented the usage of Fractional-Order Calculus in the Autonomous Parking control system, which has solved the high-precision control problem when the Autonomous Vehicle is in low speed.2 In order to overcome the problem in applying‘Long Memory Method’or‘Short Memory Method’to solve the Fractional-Order Calculus equation, the method based on Gauss Kernel to solve the Fractional-Order Calculus equation for control system is proposed. It is shown that this method has the advantage of fast calculation speed and high accuracy.3 In optimizing the parameters of the Fractional-Order Controller, a self-tuning PIλDμcontroller based on approximate policy iteration is proposed. In this algorithm, the basis function is automatically selected in accordance with the Bellman residual error, leading to an improvement of the Value Function Approximator both in approximation precision and generalization ability.4 The three-layered hierarchical intelligent control framework based on human-machine cooperation is proposed. This framework introduced human wisdom into each layer of the control system. It comprehensively utilizes human’s characteristics in qualitative perception, decision-making, planning and machine’s superiority in quantitative sensing, computing, high-precision operating. Making use of each side’s advantages, the safety and environment adaptability of the autonomous parking system is improved.5 In dealing with various working conditions of lateral parking, we have proposed the method to generate collision-free space in parallel parking. This method can easily calculate the collision-free parking area. Based on the differential flat theory and spline theory, the generation method of autonomous parallel parking trajectory with kinodynamic constraints is proposed, which can generate a viable trajectory for autonomous-parking in the collision-free space. In-vehicle experiments demonstrate that this method could not only avoid obstacles effectively, but also solve the problem of stopping to turning the steering wheel in the parking process, satisfying the demand of continuity.Through large amount of parallel parking experiments, the goal of autonomous parallel parking has been achieved, and the feasibility and effectiveness of the above-mentioned algorithms have been verified. The research results in this thesis have made a great contribution to the accomplishment of the project“Hongqi HQ3 Autonomous Parking Technology”.更多还原