Prediction of stopping maneuver considering driver´s state

This paper presents how the difference of driving patterns affects prediction of driver´s stopping maneuver. Our proposed method has two different driving models based on a driver´s state, the model in normal state and the model in hasty state. One of the models is selected depending on a driver´s state, and the maneuver is predicted in the selected model. The hidden Markov model (HMM), one of the specific form of dynamic bayesian networks (DBN), was applied for modeling of each driving pattern. For estimation of driver´s state, heart rate variability (HRV) based index was used because the index reflects stress level which seems to relate to hasty state. We have carried out a driving experiment with a simulator. As the result of evaluating the pattern in hasty state, the proposed method predicted stopping maneuver earlier than a single model method. To predict driver´s maneuver is essential to improve the human-machine interface n driver assistance technologies including a collision-warning system. Our approach is intended to enable more accurate prediction of the maneuver for these systems