改进LS-SVR在线控转向系统容错控制中的应用

Improved LS-SVR algorithm and its application on fault-tolerant control of steer-by-wire automobile

针对现阶段容错技术中对于不可直接测量变量往往采用易受扰动影响的观测器这一缺点,提出一种基于鱼群算法优化的最小二乘支持向量回归机(LS-SVR)方法用于代替传统的观测器。该方法利用鱼群算法迭代求解LS-SVR中出现的矩阵方程,从而避免了矩阵求逆过程,减少了LS-SVR算法的训练时间,并且能取得最优解。将LS-SVR应用于容错控制中的质心侧偏角估计,一个训练好的LS-SVR包含了质心侧偏角的冗余信息,可以代替观测器进行估计输出。通过仿真实验表明,所提方法收敛速度快,抗干扰能力强,效果明显提升。

In view of the disadvantage of the susceptible observer in current fault-tolerant control techniques, this paper presents the Least Squares Support Vector Regression（LS-SVR） based on fish-swarm algorithm to replace the traditional observer. In this approach, the fish-swarm algorithm is utilized to solve the matrix equation in the LS-SVR by iteration, thus, solving the inverse matrix is avoided, so the training time of the LS-SVR is reduced when the optimal solution is obtained. The LS-SVR is utilized to estimate the side-slip angle in the fault-tolerant control. The redundant information of the side-slip angle is contained in the well trained LS-SVR. It can replace the observer to output the estimate value. The simulation results show that the proposed approach provides a higher convergence speed and anti-interference ability;