Learning Based $Hinfty$ Path Following Controller Design for Autonomous Ground Vehicles Subject to S

In order to decrease computation loads for path following control of autonomous ground vehicles (AGVs), in this article, we aim to design an output-feedback path following controller by using a learning-based least squares support vector machine (LS-SVM) model. The signal transmission delay, which would deteriorates the control performance, is considered as well. First, the LS-SVM model for vehicle path following system is trained. Then, the output-feedback controller subject to stochastic communication delay, which meets the stability condition, is designed for the LS-SVM model. The modeling error and bias terms of LS-SVM model are regarded as the disturbance term. In order to attenuate the effect of disturbance, H∞ controller is designed for the AGVs to track the target path. Grey wolf optimizer is employed to solve the controller design problem with H∞ performance and transform stability condition from bilinear matrix inequality to linear matrix inequality (LMI). Then, we employ efficient LMI toolbox to solve the LMI. To guarantee the effectiveness of proposed controller, characteristics of steering wheel in a hardware-in-loop platform are studied and rate of steering angle is regarded as the actuator constraints for designing controller. Experimental results based on the hardware-in-loop platform have verified the effectiveness of the proposed control strategy.