• Journal of the Korea Society of Computer and Information
• /
• v.29 no.11
• /
• pp.57-66
• /
• 2024

In this paper, an approach based on deep learning and parameter dependent control is proposed for electronic throttle body(ETB) control which has variable parameters and nonlinear torques. Firstly we present parameter estimation method for ETB system using deep neural network. To estimate parameters of ETB, we design deep neural networks and train by use time response characteristic such as rise time, overshoot and settling time. Parameters of ETB are estimated through trained neural networks by using time response data. Secondly we design parameter dependent PID controller which is adjusted automatically with the estimated system parameter of ETB. To design optimal parameter dependent gain of PID controller, we use ITAE(Integral of time multiplied by absolute error) criteria. In addition, we design feed-forward controller to reject nonlinear torque. Finally we present simulation results of ETB syatem with parameter variation and nonlinear torque to verify controller design method.

• The Transactions of the Korean Institute of Electrical Engineers D
• /
• v.55 no.1
• /
• pp.20-23
• /
• 2006

This paper presents a new approach to the problem of designing a cascaded three-parameters controller for a given linear time invariant (LTI) plant in unity feedback system. We consider a proportional-integral-derivative (PID) and a first-order controller with the specified overshoot and settling time. This problem is difficult to solve because there may be no analytical solution due to the use of low-order controller. Furthermore, the zeros of controller just appear in the zeros of feedback system. The key idea of our method is to impose a constraint on the controller parameters so that the zeros of resulting controller are distant from the dominant pole of closed-loop system to the left as far as the given interval. Two methods realizing the idea are suggested. We have employed the characteristic ratio assignment (CRA) in order to deal with the time response specifications. It is noted that the proposed methods are accomplished only in parameter space. Several illustrative examples are given.