Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 28, 2023 - July 01, 2023
Accurate prediction of the traffic participant behavior is essential for safe and smooth path planning of autonomous vehicles. However, precise behavior prediction of surrounding dynamic objects in urban environments is still a challenging demand due to computation time and difficulties. This study proposes a model for predicting the feasibility of driving on their route of road users in dense traffic flow using the probability of driving continuation. Probability represents the difference between actual and expected behaviors by considering the interaction of traffic participants, traffic rules, and road geometry. An evaluation was conducted to check the probability of the target vehicle turning left on a straight route. The experimental results have verified the robustness of the proposed method to precisely predict the target vehicle behavior within 15m from certain reference points.
