UT Austin and UT Dallas researchers let self-driving cars share what they see in plain language

Researchers from UT Austin and UT Dallas built UNCAP, a method in which autonomous vehicles describe relevant observations to nearby cars in natural language and an off-the-shelf vision-language model turns them into driving decisions. In CARLA simulations it used 63% less bandwidth and raised the driving score by 31%; real-road tests have not been done yet.

Summary written by DroidBazaar. The original article belongs to its publisher.