The Bad PlaceEnglish – The Conversation | Self-driving cars struggle to see at night or in fog – but imitating the human brain can make them safe by Pablo Hernández Cámara, Profesor e investigador. Departamento de Ingeniería Electrónica & Laboratorio de Procesado de Imágenes, Universitat de València, Universitat de València
AI generated summary, Read the full article for complete information.
Self‑driving cars work well in clear daylight but become almost blind in darkness, rain or fog, because current AI vision systems lack the adaptive mechanisms that human eyes use. Researchers at the University of Valencia mimicked the brain’s “divisive normalisation”—a neuronal “volume‑control” that amplifies weak signals in dark scenes and attenuates bright ones—to modify standard AI models. Tests with real‑world European driving data, night‑time images from Switzerland and virtual simulators showed that the brain‑inspired models retained accurate object detection under fog and complete darkness, outperforming unmodified AI by more than 20 %. The study suggests that improving autonomous‑vehicle safety does not require larger computers or massive datasets, but rather can be achieved by borrowing evolution‑tested strategies from human vision, making AI systems more robust, adaptable, and trustworthy in all weather conditions.
#UniversityofValencia #Selfdrivingcars #AIvision #Neuralnetwork #Humanbrain #Divisivenormalisation #Switzerland #Europeandatasets #Autonomousvehicles #Braininspired #
UniversidadxClima
Prof. R. I. Sujith
AD19-Valencia2026
Jude Jackson
SciPost Physics