SITUATIONAL AWARENESS ASSESSMENT IN AUTONOMOUS VEHICLE TESTING PROFESSIONALS

Abstract

This work assesses the situational awareness (SA) capabilities of professionals actively involved in the testing of autonomous vehicles (AVs), an area where human supervisors need to manage complex systems that require interpretation and intervention. With the progression of AV technologies, the role of the tester has become more complex as they need to manage supervisory control, real-time decisions, and system overrides simultaneously. The traditional methods of assessing SA have overlooked the cognitive processes and real-time challenges during the work. To fill this gap, our study uses a multimodal neural evaluation framework that combines cognitive and physiological tools (electroencephalography (EEG), eye-tracking, and heart rate variability (HRV)) with behavioral measures of performance and workload evaluation (NASA-TLX). The test subjects were professional AV testers who were performing simulated driving exercises at three levels of task automation: low, medium, and full. The findings demonstrated a decline in awareness of the situation as automation levels increased. While the lower and medium levels of task automation sustained active engagement and accurate situational monitoring, higher automation levels elicited cognitive disengagement and sluggish response times to intervention, as well as reduction in hazard detection—regardless of low subjective workload. EEG data indicated theta and diminished beta rhythms, which are markers of mental fatigue, while eye-tracking showed diminished attention to important dials during high automation. The heart rate variability data pointed to increased stress during ambiguous and failure-prone situations, a phenomenon most pronounced among testers with lower automation levels.

The research confirms that automation does not reduce cognitive load for a user. Rather. it reallocates shifts in attention in a manner that could lower situational awareness. More experienced testers had better attention regulation, gaze consistency, and faster override decision responses which highlights the need for neurocognitive preparedness and training. For these findings, we suggest a real-time situational awareness monitoring and interface adjustment framework which physiologically adapt through biometric data streams to sustain vigilance and trust in the system. This research demonstrates critical industrial relevance, as anthropometric AV test environments need to prioritize attitudinal human factors to sustain cognitive engagement. Incorporating adaptable interfaces, tailored training, and biometric real-time monitoring can enhance tester performance and safety. Ethically, privacy, inclusivity, and data transparency must be prioritized in these frameworks. This research provides a starting point for advanced cognitive integrated AV systems, which achieves the efficiency of automation while providing enduring human situational awareness and responsive automation aids.