Chinese researchers have developed a system that uses passengers' brain signals to improve autonomous vehicle safety in risky situations. The technology, published in Cyborg and Bionic Systems on May 13, 2025, employs functional Near-Infrared Spectroscopy (fNIRS) to monitor stress and risk perception in real time.
The system from Tsinghua University combines brain data with autonomous driving software through a deep reinforcement learning algorithm. When passengers show elevated stress levels, vehicles automatically shift to more conservative driving modes. This approach outperformed traditional methods in learning speed, safety, and comfort during testing.
Lead researcher Xiaofei Zhang stated the fNIRS technology provides cognitive information about human risk perception and emotional states. "Our study introduces an intelligent decision-making algorithm based on fNIRS by analyzing passengers' physiological states," Zhang said in a news release. The algorithm aims to improve safety and decision-making efficiency when autonomous vehicles face risky scenarios.
Current limitations include relatively simple driving scenarios tested and participants from narrow demographic ranges. The researchers acknowledge findings may not apply to all real-world situations. Future work will validate the algorithm in more complex scenarios and integrate vehicle sensor data for enhanced risk assessment.
The research arrives as companies like Tesla continue developing hands-free driving systems that have faced scrutiny following crashes. Traditional autonomous systems still struggle with fast-changing, high-risk situations where human intuition could provide critical safety inputs.
Brain-monitoring technology represents a novel approach to bridging the gap between human perception and machine decision-making in autonomous vehicles. The non-invasive fNIRS method tracks brain activity linked to stress, emotions, and risk perception without physical intervention.
Industry analysts note this research could influence next-generation safety systems, potentially creating hybrid approaches that combine human cognitive feedback with machine learning algorithms. However, widespread implementation would require addressing privacy concerns and developing robust, diverse testing protocols.
The study's publication in a peer-reviewed journal suggests academic validation of the core concept, though commercial applications remain years away. Automotive manufacturers and tech companies developing autonomous systems are likely monitoring such research as they seek competitive advantages in safety features.
As autonomous vehicle technology advances, integrating human feedback mechanisms could become increasingly important for public acceptance and regulatory approval. Systems that can detect and respond to passenger discomfort may address one of the key psychological barriers to widespread adoption of self-driving cars, according to the Union of Concerned Scientists.
