Driver State Modeling through Latent Variable State Space Framework in the Wild

Arash Tavakoli; Steven Boker; Arsalan Heydarian · 2022 · arXiv

URL: http://arxiv.org/abs/2203.00834v1

archive: archived pipeline: cataloged verified

Abstract

Analyzing the impact of the environment on drivers' stress level and workload is of high importance for designing human-centered driver-vehicle interaction systems and to ultimately help build a safer driving experience. However, driver's state, including stress level and workload, are psychological constructs that cannot be measured on their own and should be estimated through sensor measurements such as psychophysiological measures. We propose using a latent-variable state-space modeling frame

Summary

In-the-wild multimodal driving study (Tavakoli, Boker, Heydarian) of N=11 participants, treating driver stress and workload as latent psychological constructs estimated jointly via a latent-variable state-space model from heart rate, gaze variability, and facial action-unit intensity. External context (traffic density via vehicle count, secondary task demands) perturbs the latent states, and previous-timestep latent states are highly associated with current states. Individual differences in how perturbations propagate to stress and workload are documented, and the framework analyzes lag between stress and workload as a marker of within-driver psychophysiological information transfer.

Key finding

A latent-variable state-space framework jointly recovers stress and workload from multimodal driver psychophysiology in naturalistic driving, exposing individual-difference structure and stress-workload temporal lags that single-channel models cannot capture.

Methodology

Exp 1: 10 participants, repeated measures across 6 sessions from 26 total. Exp 2: 20 participants, Old/New sequence comparison. On-road driving paradigm with DRT and NASA-TLX measures.

Sample size: Exp 1: N=10; Exp 2: N=20

Quality score: 5 / 5

Topics