Sony announced on November 8 the global debut of a visual language model based driving takeover hazard warning technology
Sony uses a visual language model to analyze road imagery ahead of the vehicle and judge whether conditions approach or exceed the driving system design domain, providing three levels of handover prompts.
On November 8, it was reported that on November 5, 2026, Sony China unveiled at the ninth China International Import Expo the global debut of a mobile mobility technology for driving takeover hazard warnings based on visual language models. The technology analyzes road imagery ahead of the vehicle with a visual language model and judges whether road conditions approach or exceed the design operating domain of the driver assistance system, giving drivers graded prompts so they can understand the situation and prepare a response.
Takeover prompts come in three levels: a reminder to prepare for takeover when approaching the boundary, guidance to take over actively when conditions exceed the range but are not yet urgent, and an emergency takeover prompt for sudden obstacles or perception failures. During development, the team used digital twin simulation environments to generate data for daytime, nighttime, heavy rain and high-speed driving conditions for model training and validation. The technology has already begun real road testing with Tsinghua University's School of Vehicle Engineering in Xiangcheng District, Suzhou, supporting automakers and driver assistance vendors in handover timing decisions and human-machine interaction design.