Synthesis and Approximation of Control in Adaptive Cruise Control Systems of Commercial Vehicles

Abstract

In this paper, we consider the problem of the development of an algorithm of the adaptive cruise control functioning operating in the conditions of powertrain gear ratio varying in a wide range and vehicle velocity changing. The functioning of a classical cruise control system is generally based on the usage of a PID-controller with constant coefficients. However, despite the easiness of its tuning and physical realization and also its relatively high robustness this class of control devices cannot guarantee the cruise control system optimal functioning in all driving conditions because the plant is not timeinvariant and linear. To overcome the above shortcomings, in this research we consider the possibility of neural network realization of a commercial vehicle adaptive cruise control algorithm.In this paper, we propose the mathematical model of a commercial vehicle longitudinal motion designed for the control system analysis and synthesis. We carry out the PI-controller coefficients tuning to control the vehicle longitudinal velocity in various driving conditions of a commercial vehicle. We show that the controller coefficients vary according to a rather complex law. Therefore, we propose the algorithm of the adaptive cruise control functioning based on the approximation of the controller coefficients by the artificial neural network. The network used is the multilayer perceptron and it has ten neurons in the hidden layer to provide the high quality of the approximation. We carry out the training of the neural network by the Levenberg-Marquardt method with a sample of a total volume of 500 points, obtained using standard methods of controller synthesis. We verify the correctness of the obtained results through the computer simulations of the vehicle acceleration from 0 to 100 km/h, proving that the PI-controller coefficients, providing the required transient responses, significantly vary depending on the current state of the vehicle. The approach of the PI-controller coefficients approximation presented in this paper may be further used in the design of adaptive control systems able to function effectively in various operating modes.