Embodied reasoning becomes task orchestration
ER 2 adds video understanding, multi-step planning, progress checking and higher-level orchestration above lower-level robot control.
ER 2 adds video understanding, multi-step planning, progress checking and higher-level orchestration above lower-level robot control.
Google DeepMind expanded its robotics model from upper-body tasks to coordinated locomotion, posture changes and manipulation across embodiments.
Reported fundraising talks indicate investor appetite is concentrating around cross-embodiment robot foundation models and reusable intelligence platforms.
The local model targets rapid adaptation to new robot bodies and objects using fewer than two hundred examples.
Google separated higher-level planning from physical control while broadening multi-step task execution.
RedFlow converts failed robot actions into localized corrections for vision-language-action policy improvement.
Action-conditioned predictive models are used to improve generalizable embodied decision-making.
German researchers evaluated lightweight personalized memory for sustained human-humanoid interaction.
The research introduces temporally constrained latent-action learning for more coherent embodied control.
An AMD-accelerated stack spans data-center training, simulation and Ryzen AI edge deployment for VLA manipulation.