Study on Behavior Prediction Using Multi-object Recognition and Map Information in Road Environment

Abstract

Autonomous vehicle technology offers the possibility of significant benefits to social welfare such as reducing crashes, congestion, fuel consumption, and so on. Among them, collision prediction and avoidance system are especially important topics in real road scenes. Here we introduce the novel behavior prediction system based on multi-object recognition and map information. In multi-object recognition, it is necessary to recognize the object on the road with high accuracy and speed. Therefore, we perform a method to apply Variable Window Approach to 2DHOG feature extraction method. In predicting behavior, the distance relationship between the host vehicle and the obstacle is important. We will combine the digital map and the map obtained from Visual SLAM to clarify the relationship between these distances. Finally, we verify the usefulness of this system by multi-object recognition experiment and behavior prediction experiment.