基于神经网络逆系统的智能汽车纵横向解耦控制

针对汽车纵横向运动中的耦合现象,以四轮驱动、前轮转向的智能汽车为研究对象,建立汽车纵横向动力学模型并通过Interactor算法对模型的可逆性进行分析.在已有的传统伪线性系统结构的基础上,根据智能汽车的特点,建立了可对接智能汽车上层规划模块的伪线性系统.为了实现汽车纵横向运动之间的解耦,采用基于神经网络逆系统的解耦控制策略,构造神经网络并对其进行训练,并将神经网络逆系统与内模控制器组成闭环控制回路,对纵向速度和横摆角速度进行内模反馈调节,进一步提升控制系统的性能.仿真结果表明,所设计的基于神经网络逆系统的控制方法能实现良好的解耦特性,且相比于其他的控制方法,在各种输入条件下,都能实现对于期望速度和期望横摆角速度良好的跟踪性能,同时,质心侧偏角始终被控制在一个较小的范围内,这有利于智能汽车路径跟踪的精确性和行驶稳定性.

Decoupling Control of Longitudinal and Lateral Motion for Intelligent Vehicle Based on Neural Network Inverse Method

Aiming at the coupling phenomenon of longitudinal and lateral motions for automobile, autonomous vehicle with Four-wheel-driving and front-wheel-steering was set as the subject investigated. A dynamic model which reflects the longitudinal and lateral motion of vehicle was established and the reversibility of this model was analyzed by the interactor algorithm. On the basis of the existing classical structure of pseudo-linear system, the pseudo linear composition system with the ability to fit the upper level planning system of intelligent vehicle was established according to the characteristic of intelligent vehicle. In order to realize the decoupling of longitudinal and lateral motions for vehicle, an approach based on network inverse method was proposed as the decoupling control strategy in this paper, which can be combined with the internal model controller to form closed loop structure and it can significantly improve the performance of the plant by feedback and adjust the longitudinal speed and yaw rate of automobile. The simulation results validated the decoupling performance of the proposed approach. The results also showed that when compared with other control algorithms, the proposed approach can achieve good tracking performance of longitudinal speed and yaw rate under varieties of input condition. Further, the sideslip was constrained in a small range, which is beneficial to the path tracing accuracy and the stability of autonomous vehicle.