다른 예측 주기를 이용한 모델예측 기반 횡방향 조향 제어

Abstract

In this dissertation, we present how model predictive control (MPC) is applied to a lane keeping system and how real-time solution is calculated. In the case of fast systems, real-time solution is very difficult to obtain in the sample time because the system has complex formulations involving many constraints. And, if prediction horizon is long, the system can get optimal solution while assuring closed-loop system stability. The formulations of MPC is, however, very complex as increase of prediction horizon. On other hand, if prediction horizon is very short, computation overhead is reduced but stability and performance of the closed-loop system are not guaranteed. To solve the problems, we propose the three methods such as dual prediction horizons MPC, proximate MPC and fast MPC. The dual prediction horizons MPC method reduces the length of prediction horizon as well as maintains the stability and performance. Optimal solution of the proximate MPC is calculated by interpolating pre-optimal solutions regarding vertices. And the fast MPC decreases the number of constraints using weighted forgetting factors in order to find effective active constraints. The comparison of performance of the proposed control methods are validated via simulations implemented in real-time environment.