Self-driving cars struggle to see at night or in fog – but imitating the human brain can make them safe
Publish Date: 2026-05-12 05:47:00
Source Domain: theconversation.com
- Current AI vision systems in self-driving cars function well under good visibility conditions but struggle in adverse weather like fog, darkness, or rain, posing significant risks.
- Research from the University of Valencia suggests mimicking human vision’s adaptability as a solution, specifically through the use of divisive normalization, a biological mechanism that adjusts signal strength.
- By incorporating a brain-inspired “volume control” mechanism in AI models, the researchers’ simulations showed that modified AI performed over 20% better in adverse conditions compared to standard AI models.
- The enhanced AI, which learns from nature, showed improved adaptability and safety in diverse and challenging driving scenarios, highlighting the potential of biological inspiration to improve AI robustness.
- The study emphasizes that the key to safer, more adaptable AI may lie in understanding and emulating biological processes, rather than increasing computing power or data collection.
