도로 기하 정보를 적용한 다중 모델 기반의 확률적인 차량 경로 예측

Abstract

This paper proposed a multi-model based stochastic vehicle trajectory prediction algorithm using road geometry information. The probability of collision with surrounding vehicles has been suggested as an indicator for determining the suitability of an autonomous driving system. To estimate vehicle behavior in complicated situation in usual roadway, the road geometry and the model-based estimation of various vehicle behaviors should be integrated. the Interacting Multiple Model (IMM) filter with a curvilinear coordinate system is used for model integration to estimation vehicle behavior from complex road geometry information. Moreover, each probabilities of basis models are used to predict trajectory of surrounding vehicles. This proposed method was designed in simulation environment and validated with various scenarios such as lane keeping and lane changing in curved road.