Data-driven Design of Model Error Compensator and Fictitious Reference Signals for Vehicle Velocity Control of Autonomous Driving

Vehicle velocity control has attracted significant research attention for autonomous driving applications. To achieve reliable autonomous driving, the desired velocity control should be realized. However, it is challenging to realize the desired control response because it is difficult to identify the accurate vehicle model. A feedback controller based on the nominal model cannot realize the desired velocity controls when the nominal error is extremely large. In this study, we propose a velocity control based on a model error compensator (MEC) and fictitious reference iterative tuning (FRIT). The proposed method can calculate the control signal to minimize the relative error between the closed-loop system and the normative model. By designing the tunable parameters of the MEC in the framework of FRIT, the desired control response can be achieved using one-shot experimental data. We verified the proposed method using a vehicle dynamics blockset.