情報処理学会 第81回全国大会講演要旨

This paper presents a novel autonomous driving method that can control the own vehicle while predicting mutually cooperative behaviors of others, which have not been addressed by any previously methods. The method is characterized by three features: (1) path search in position-velocity space which makes it possible to optimize motions under constraints in terms of acceleration, (2) the objective function which incorporates vehicle-obstacle/vehicle-vehicle relationships, and (3) no learning which leads to unlimited domains/situations. Simulation experiments showed that the method could produce appropriate motions even in complicated situations where one vehicle had to give way to the other vehicles.