Abstraction and Implementation of Human Skill by Hybrid Dynamical System Theory — Application to an Automatic Driving System —

抄録

In this paper, we propose a method to abstract human manipulation skill, which can be useful in systems of machine intelligence and human operator assistance. It is, however, not easy to model human operation by traditional methods due to the lack of good physical models for human skill, In our approach, human manipulation skill is considered as a hybrid dynamical system (HDS). Specifically, a hinging hyperplane autoregressive exogenous (HHARX) model is employed as it is able to deal with manipulation modes and switches among them simultaneously. Such an HHARX model can be obtained by system identification via mixed-integer linear programming (MILP). As a typical example, we apply our approach to an automatic driving system of a small radio-controlled vehicle. Both simulation and experimental results illustrate the effectiveness of the proposed method.