Live demo failure highlights communications resilience
A humanoid demonstration reportedly entered a safety shutdown after a communication fault, spotlighting fail-safe behavior and integration quality.
Semiconductor Domain: real-time control
Gemini Moves From Manipulation to Whole-Body Control
Gemini Robotics 2 reporting extends control toward locomotion, dexterity and coordinated whole-body tasks.
VLA Failures Become Corrective Training Signals
RedFlow converts failed robot actions into localized corrections for vision-language-action policy improvement.
Temporal Constraints Target More Stable Latent Actions
The research introduces temporally constrained latent-action learning for more coherent embodied control.
SymmGrid accelerates direct on-robot learning for contact tasks
Symmetry-based trajectory augmentation improved training speed and success rates on real robot insertion, routing, and relocation tasks.
Gemini Robotics 2 Reaches Whole-Body Control
Gemini Robotics 2 extends control toward walking, crouching, dexterous manipulation and coordinated whole-body tasks.
Embodied Reasoning Expands to Longer Tasks
Google broadened dexterity, multi-step execution and the separation of higher-level planning from physical control.
DLAM Adds Temporal Constraints to Latent Robot Actions
DLAM introduces temporal constraints intended to make latent robot actions more coherent across sequential control decisions.
Google DeepMind releases Gemini Robotics 2
Gemini Robotics 2 expands robot planning, dexterity, and full-body control across humanoid and task-oriented platforms.
RedFlow turns VLA failures into action corrections
RedFlow redirects detected failures into action-level corrections for flow-matching vision-language-action policies.
