Autonomous Vehicle Trajectory Following: A Flatness Model Predictive Control Approach With Hardware-in-the-Loop Verification
Zejiang Wang, Jingqiang Zha, Junmin Wang
Abstract
Trajectory following of autonomous vehicle is a challenging task because of the multiple constraints imposed on the plant. Therefore, Model Predictive Control (MPC) is becoming prevail in vehicle motion control as it can explicitly handle system constraints. However, MPC, grounded in real-time iterative optimization, entails a considerable computational burden for current electronic control units. To mitigate the MPC execution load, a popular strategy is to linearize the original (nonlinear) system around the current working point and then design a Linear Time-Varying MPC (LTVMPC). Nevertheless, the successive linearization introduces extra modeling errors, which may impair the control performance. Indeed, if the plant model satisfies the `differential flatness' condition, it can be exactly linearized to the Brunovsky's canonical form. In contrast to the LTV model, this newly appeared linear form reserves all the nonlinear features of the native plant model. Based on this equivalent linear system, a Flatness Model Predictive Controller (FMPC) can be formulated. FMPC on the one hand, improves the control performance over an LTVMPC because it avoids extra modeling errors from the local linearization. On the other hand, it entails a much lighter computational load versus a nonlinear MPC thanks to its linear nature. Real-time simulations conducted on a hardware-in-the-loop system indicate the advantages of the proposed FMPC in autonomous vehicle trajectory following.