Autonomous lane tracking reflecting skilled/un-skilled driving characteristics

This paper presents an autonomous lane tracking that reflects different driving characteristics using model predictive control (MPC). We consider that human driver minimizes a cost function depending on his/her skill, experience, and preference on driving. The cost function of MPC can be used to model personal driving characteristics. To identify the parameters of cost function, at first, we analyze the difference of driving characteristics between skilled and un-skilled drivers through experiments of driving behavior on a driving simulator (DS). Next, we introduce "meta-performance indices" that can evaluate human driving data of experiment and results of autonomous driving. These parameters are expected to express the driving characteristics as a group, i.e., not tuned to personal driver. Finally, the validity of the proposed system is verified.