Robust visual feedback control design using driving data on straight and curved courses

This paper presents a modeling approach to controlling a vehicle through visual feedback control as well as through H_∞ control. Two models are designed using closed-loop identification. The usefulness of closed-loop subspace identification for the camera-equipped vehicle in this study is discussed. Additionally, the most useful vehicle identification using visual feedback controls is identified as MOESP-type1 closed-loop subspace model identification [1]. A fitting rate is used to evaluate the model effectiveness. A robust controller is then designed on the basis of the constructed models. More specifically, a model error for each model is formulated for curved and straight paths using an H_∞ control algorithm for the vehicle controller while remaining within a given error. A nominal plant is regarded as the model given by the data for the straight path. The model error is estimated as the difference in the parameters of the models given by the straight and curved paths. The validity of H_∞ control is discussed using the model errors from simulation and experimental results.