Modeling Pipeline Driving Behaviors: A Hidden Markov Model Approach

Abstract

Driving behaviors at intersection are complex because drivers have to perceive more traffic events than normal road driving and thus are exposed to more errors with safety consequences. Drivers make real-time responsesin a stochastic manner. This paper presents our study using Hidden Markov Models (HMM) to model driving behaviors at intersections. Observed vehicle movement data are used to build up the model. A single HMM is used to cluster the vehicle movements when they are close to intersection. The re-estimated clustered HMMs provide better prediction of the vehicle movements compared to traditional car-following models. Only through vehicles on major roads are considered in this paper.