智能汽车故障诊断与容错控制关键技术研究

Intelligent autonomous vehicles are recognized as one option of addressing traffic jams and enhancing driving safety and riding comfort, while fault detection and diagnosis (FDD) and fault-tolerant control (FTC) play a profoundly important role in intelligentizing vehicles. The focuses on FDD and FTC for longitudinal-lateral control systems of intelligent autonomous vehicles are placed in this project, such that the safety of autonomous driving can be ensured. Upon the accomplishment of this project, systematic theory and methodologies will be developed. More specifically, the research to be conducted includes 1) model and learning mechanism based FDD for the actuators, sensors, and other key components of the intelligent vehicle, which is different from traditional FDD methods; 2) FTC with explicit consideration of actuator amplitude and rate limits under faults, disturbances, couplings, and model uncertainties; 3) finite-time FTC in the case of the very limited time for fault accommodation; and 4) integration of all the subsystems and validation through experimental tests. In summary, this project aims at addressing the scientific issues for FDD and FTC during the autonomous driving of intelligent autonomous vehicles. Moreover, the accomplishment of this research project will pave the solid foundation for the driving safety of intelligent autonomous vehicles from both theoretical and technical perspectives.

智能汽车是解决交通拥堵、提高汽车行驶安全性和乘员舒适性的有效途径之一，而故障诊断与容错控制是实现汽车智能化的核心支撑技术。本项目面向智能汽车自主行驶的高安全性需求，以智能汽车为对象，研究智能汽车部件故障诊断方法和纵横向容错控制策略。具体内容包括：1) 针对难以精确建立智能汽车部件数学模型的难题，探索基于模型和自学习机制的故障诊断新理论和新方法；2) 考虑智能汽车外部干扰、纵横向耦合与模型不确定性因素，针对智能汽车部件故障，研究执行器输出幅值和速率受限的纵横向容错控制新策略；3) 充分考虑故障调节时间极其有限这一安全性相关的重要问题，研究基于有限时间收敛的纵横向容错控制新方法；4) 实现智能汽车容错控制系统的集成及实验验证。本项目旨在解决智能汽车的故障诊断与容错控制等基础科学问题，为实现智能汽车安全行驶提供理论基础与技术支撑。

智能无人汽车是解决交通拥堵、提高汽车行驶安全性和乘员舒适性的有效途径之一，国务院发布的“十三五”文件把智能无人汽车列为国家战略性新兴产业，具有重大的研究价值。本项目针对智能无人汽车故障诊断与容错控制问题，以基于模型和自学习机制理论为基础，考虑到执行器固有的饱和限制，提出了快响应的智能汽车容错控制方法。主要贡献如下：（1）在充分考虑空气作用力等复杂因素的基础上，建立智能汽车纵横向动力学模型，并分析纵横向耦合、系统关键部件故障情况下控制的可行解空间；（2）在智能汽车关键部件故障模态分析基础上，依据各部件的物理机理和机械特性，建立正常条件和故障条件下的数学模型；（3）基于智能汽车各关键部件初步模型和离线、在线数据，利用自学习机制优化模型，提出了一种具备较强鲁棒性和较高准确度的故障诊断方案；（4）利用故障诊断信息，结合智能汽车执行器的实际能力，提出了具有快响应特点的容错控制技术；（5）在实验室无人车平台上验证了本项目所提出算法的有效性，研究结果发表SCI论文14篇（标注本项目资助号）。