Embodied Reasoning for Robotics

3 posts · Apr 2026 – Jul 2026

What this trail covers

Gemini Robotics ER 1.6 enhances embodied reasoning with improved spatial reasoning, multi-view understanding, and instrument reading capabilities. It builds upon previous versions by offering more precise object detection, counting, and success detection, along with the ability to read complex gauges and sight glasses. The model also demonstrates enhanced safety compliance and adherence to physical constraints. It is available via the Gemini API and Google AI Studio.

All DeepMind trails

Jul 2026Gemini Robotics 2 brings whole body intelligence to robotsThis post introduces Gemini Robotics 2, which significantly advances embodied reasoning for robots. Key contributions include: 1. Whole-body control for humanoids (e.g., walking, crouching, manipulating objects) enabling tasks like cleaning cluttered rooms. 2. Advanced dexterity for hands and grippers, demonstrated by tying knots, sealing ziplock bags, and complex packing tasks. 3. Enhanced embodied reasoning (ER 2) for longer task sequences (minutes, hundreds of decisions), improved progress understanding, and multi-robot collaboration. 4. Fast on-device adaptation (hours, <200 examples) for new robot embodiments with Gemini Robotics On-Device 2. 5. Improved safety features with Gemini Robotics ER 2, including better human detection, refusal of unsafe tool calls, and proactive human intervention requests, benchmarked by ASIMOV-Agentic.
Jul 2026Introducing Gemini Robotics ER 2Gemini Robotics ER 2 introduces advancements in continuous progress classification and precision moment-finding for temporal intelligence, enabling robots to track task progress and identify critical event frames. It also enables multi-robot collaboration for complex task handoffs. Spatial reasoning is improved with success/failure detection on raw video feeds, general instrument reading across various display types, and enhanced spatial VQA. Safety is advanced with improved Safety Instruction Following and Human Proximity benchmarks, and a new benchmark for evaluating safe VLA orchestrators.
Apr 2026Gemini Robotics ER 1.6: Enhanced Embodied ReasoningGemini Robotics-ER 1.6 introduces significant improvements in spatial reasoning, multi-view understanding, and instrument reading for robots. Key advancements include enhanced pointing capabilities for precise object detection, counting, and relational logic; improved success detection for greater autonomy; and the novel ability to read analog gauges and sight glasses, a capability developed in collaboration with Boston Dynamics. The model also shows superior compliance with Gemini safety policies and physical safety constraints, and improved accuracy in identifying safety hazards in text and video scenarios.
Advertisement