银河通用发布端到端导航大模型,联合宇树机器狗完成长程验证
news flash·2025-06-01 01:59
Core Viewpoint - The company has launched an end-to-end navigation model called TrackVLA, which features capabilities such as pure visual environment perception, natural language instruction driving, and end-to-end output of language and robotic actions [1] Group 1 - The TrackVLA model has been developed independently by the company [1] - The model has been successfully integrated into a robotic dog named "Er Bao Bodyguard" for child supervision [1] - The model has undergone rigorous long-range validation in real-world scenarios without prior training [1]