Driver behavior modeling near intersections using Hidden Markov Model based on genetic algorithm

Abstract

Driver behavior modeling plays a significant role in the development of Advanced Driver Assistance Systems (ADAS) for assisting drivers in different driving scenarios. One of the scenarios where high numbers of traffic accidents occur is road intersection. It is vital to develop driver behavior models near intersections in order for the ADAS to plan a proper action in avoiding accidents. In this paper, Hidden Markov Models (HMMs) for driver behavior near intersections are trained using Genetic Algorithm combined with Baum-Welch Algorithm based on the hybrid-state system (HSS) framework. HMM is usually trained using Baum-Welch which is easily trapped at local maxima. GA solves this problem by searching the entire solution space. Consequently, the best driver behavior model is trained. In the HSS framework, the vehicle dynamics are represented as a continuous-state system (CSS) and the decisions of the driver are represented as a discrete-state system (DSS). The continuous observations from the vehicle, such as acceleration, velocity and yaw-rate, are used by the proposed technique to estimate the driver's intention at each time step. The models are trained and tested using naturalistic driving data obtained from the Ohio State University, in an experiment with a sensor-equipped vehicle that was driven in the streets of Columbus, OH. The proposed framework improves the HMM accuracy in estimating the driver's intention when approaching an intersection with over 10% higher accuracy.