Driving risk cognition of passengers in highly automated driving based on the prefrontal cortex activity via fNIRS

Wang, Hong; Zhang, Xiaofei; Li, Jun; Li, Bowen; Gao, Xiaorong; Hao, Zhenmao; Fu, Junwen; Zhou, Ziyuan; Atia, Mohamed · 2023 · Crossref

DOI: 10.1038/s41598-023-41549-9

archive: archived pipeline: cataloged verified

Get this paper ↗ (DOI — opens at the source; we link to it, we don't host it)

Summary

This study addresses the safety challenges in high-level automated vehicles, specifically focusing on Safety of the Intended Functionality (SOTIF). As AI algorithms possess "black box" properties and functional deficiencies, the authors propose using brain-computer interfaces (BCI) to monitor passenger mental states rather than driver workload, since humans act as passengers in highly automated systems. The research aims to quantify passengers' driving risk cognition by analyzing prefrontal cortex activity via functional near-infrared spectroscopy (fNIRS), providing a foundation for human-centered intelligent systems that can enhance vehicle safety by integrating passenger cognitive responses into decision-making processes. The experimental design involved 20 participants (15 males, 5 females; mean age 29.45) undergoing a driving simulator experiment using hardware-in-the-loop equipment and Virtual Test Drive software. An eight-channel fNIRS device (OctaMon+) monitored cerebral oxygen exchange (△COE), derived from oxygenated and deoxygenated hemoglobin concentrations. Participants experienced four challenging scenarios: lead vehicle autonomous emergency braking, left and right lane cut-ins, and pedestrian crossing. Data were segmented into low-risk and high-risk episodes based on a calculated "risk field" indicator involving relative distance and speed. Feature extraction utilized mean values and K-SVD dictionary learning, while statistical analysis employed T-tests and Generalized Linear Models (GLM). To validate simulator findings, a secondary experiment was conducted in real-world conditions in Changsha, China, using a FAW E-HS9 automated vehicle for right-lane cut-in scenarios. The results identified Brodmann area 10 (specifically channel 8, located in the left inferior frontal gyrus) as the most active region in response to driving scenario risks. T-test analyses revealed that mental activity responses to lateral risks (lane cut-ins and pedestrian crossing) were significantly stronger than those to longitudinal risks (emergency braking). The pedestrian crossing scenario elicited the strongest neural response, correlating with the fastest change in the risk field. A positive linear correlation was established between cerebral oxygen exchange in channel 8 and the risk field using GLM. Furthermore, demographic analysis indicated that driving experience and male sex were associated with heightened sensitivity to longitudinal risks. The real-world validation confirmed the simulator results, showing significant differences in cerebral oxygen exchange between low- and high-risk episodes during right-lane cut-ins. The study concludes that passenger prefrontal cortex activity, particularly in Brodmann area 10, serves as a reliable biomarker for driving risk cognition. This finding supports the feasibility of designing human-centered intelligent systems that utilize passenger brain states to improve SOTIF in automated vehicles. By integrating passenger-in-loop decision-making via reinforcement learning, these systems could potentially compensate for AI algorithmic deficiencies, thereby enhancing overall traffic safety in high-level automated driving environments.

Provenance

The full processing record for this entry. Every stage of this paper's journey through the pipeline is logged — what ran, with which tool and model, how many attempts it took, and when it last completed.

Summary generated by qwen3.6-27b-nvidia on 2026-08-10; verification: verified.

Topics

Ranked by relevance to this paper. Hover a topic for its definition.

• situational awareness
• passenger effects
• risk taking

Information type

What kind of knowledge this paper contributes, grouped by family — independent of topic (what it is about) and method (how it was studied).

• Empirical Findings: physiological data
• Theoretical Contribution: computational model, theory or model