Gemini Robotics 2 brings whole body intelligence to robots(Gemini Robotics 2:為機器人帶來全身智能)
Gemini Robotics 2 brings whole body intelligence to robots(Gemini Robotics 2:為機器人帶來全身智能)
來源: Google DeepMind 原文日期: 2026-07-28
中文摘要
Google DeepMind 發布 Gemini Robotics 2,號稱「從腳到指尖」的全身智能(whole-body intelligence),讓機器人能推理每一個動作、完成走、蹲、伸展、抓取與清掃等多樣任務。系列包含三個模型:Gemini Robotics 2(最先進的 vision-language-action 模型,可控制整個人形機器人與雙臂機器人,具備新的雙手/夾爪靈巧操作能力)、Gemini Robotics ER 2(embodied reasoning 模型,充當 agent,可與人類溝通、理解物理世界、規劃長達數分鐘的多步任務,並支援多機器人團隊協作)、Gemini Robotics On-Device 2(最高效的本地運行 VLA,數小時內即可適應全新機器人機體,通常只需不到 200 個樣本)。安全方面推出 ASIMOV-Agentic 基準,衡量拒絕不安全 tool calls 與不確定時主動請求人類介入的能力;ER 2 宣稱是迄今最安全的機器人模型。ER 2 已上線 Google AI Studio 與 Gemini Enterprise Agent Platform(private preview)。關鍵洞察
原文關鍵句(英文保留)
> "From feet to fingertips — we are teaching robots intelligent whole-body control, fine dexterity, and teamwork to complete a broad range of complex tasks."> "Unlocking the true potential of robotics requires moving past single-task automation toward general-purpose intelligence."
> "ASIMOV-Agentic... measures the embodied reasoning agent's ability to refuse unsafe tool calls from a VLA. It also measures the agent's ability to predict whether a task is possible and to proactively request human intervention when uncertain."