Experimental investigation on developing human-centric AEB based on drivers' collision-avoidance capability in...

Area of Science:

• Human-computer interaction
• Automotive engineering
• Traffic safety research

Background:

• Advanced Driving Assistance Systems (ADAS) show limited adaptability to human drivers, causing coordination issues and increasing collision risks in critical situations.
• Current systems lack quantitative methods to assess driver collision-avoidance capabilities across varying automation levels.
• Integrating drivers' inherent collision-avoidance behaviors into ADAS design is crucial for enhancing safety.

Purpose of the Study:

• To develop quantitative methods for assessing driver collision-avoidance capabilities in safety-critical scenarios.
• To design and evaluate a capability-based Autonomous Emergency Braking (AEB) system that adapts to driver abilities.
• To improve the safety and reliability of human-vehicle shared driving systems.

Main Methods:

• Conducted a simulator-based experiment with 45 drivers across three safety-critical highway scenarios.
• Quantified driver collision-avoidance capability using naturalistic behaviors and collision rates.
• Developed a capability-based AEB concept with graded braking thresholds linked to driver capabilities.

Main Results:

• Collision risk increased by 59.7% when scenario urgency surpassed the derived driver capability boundary.
• The capability-based AEB system reduced collision rates by 17.2% and occupant injury severity by 10% compared to conventional AEB.
• Nuisance intervention rates decreased by 24.4%, enhancing operational reliability.

Conclusions:

• Adapting AEB intervention timing based on empirically derived driver capabilities is feasible and effective.
• The proposed capability-based approach enhances safety, reduces injuries, and improves the reliability of ADAS.
• This methodology offers insights for designing more human-centric ADAS and addressing takeover challenges in shared driving.