Bayesian nonparametric modeling of driver behavior

Abstract

Modern vehicles are equipped with increasingly complex sensors. These sensors generate large volumes of data that provide opportunities for modeling and analysis. Here, we are interested in exploiting this data to learn aspects of behaviors and the road network associated with individual drivers. Our dataset is collected on a standard vehicle used to commute to work and for personal trips. A Hidden Markov Model (HMM) trained on the GPS position and orientation data is utilized to compress the large amount of position information into a small amount of road segment states. Each state has a set of observations, i.e. car signals, associated with it that are quantized and modeled as draws from a Hierarchical Dirichlet Process (HDP). The inference for the topic distributions is carried out using an online variational inference algorithm. The topic distributions over joint quantized car signals characterize the driving situation in the respective road state. In a novel manner, we demonstrate how the sparsity of the personal road network of a driver in conjunction with a hierarchical topic model allows data driven predictions about destinations as well as likely road conditions.

Managing Driving Modes in Automated Driving Systems

“…This is often accomplished via vehicle-oriented (e.g., acceleration or driving path) or driver-oriented (e.g., eye closure or hand position) approaches (Hecht et al 2018, Akai et al 2019. Given the substantial effect of driver behavior on roadway safety (Brookhuis andDe Waard 2010, Wang et al 2020), predictive models have been a focus of recent DSM research (Torres, Ohashi, and Pessin 2019;Yi et al 2019b), with a few notable examples adopting a Bayesian perspective (Agamennoni, Nieto, and Nebot 2011;Straub, Zheng, and Fisher 2014). We build upon this literature by constructing an alternative Bayesian model for DSM.…”

Decision support issues in automated driving systems

“…Whether human‐ or vehicle‐oriented, probabilistic and statistical‐based models have proven effective for the purpose of characterizing driver behavior. More specifically, machine‐learning models have been a primary focus of DSM research recently (Akai et al., 2019; Torres et al., 2019; Yi et al., 2019), with a few notable examples adopting a Bayesian perspective (Agamennoni et al., 2011; Straub et al., 2014). A critical determination in these statistical‐based models are the classification levels of a driver's state.…”

Comparison of different driving style analysis approaches based on trip segmentation over GPS information

“…Straub et al in [15] proved that HDPs are able to obtain descriptive topics about road-states considering a small set of car signals. Starting from this points we used the same approach to compare driver models belonging to different drivers to allow a clustering based on their driving behaviour and habits.…”