Self-evolving embodied learning adds a criticality model
A criticality-aware framework seeks to guide self-evolving embodied learning toward safer and more consequential experiences.
Archives: News Records
Social navigation learns from human-leg sensing
A navigation approach uses human-leg sensing to improve robot behavior around people in shared spaces.
LabEvolver applies training-free experience evolution to physical agents
LabEvolver explores safe, grounded adaptation for embodied agents operating in wet-lab environments without conventional retraining.
Cross-embodiment transfer moves toward reusable behavior representations
A cross-embodiment method seeks reusable behavior representations spanning different robot bodies and control spaces.
World Action Planner links prediction directly to decisions
The World Action Planner uses action-conditioned world models to improve generalizable robot decision-making.
RoboBRIDGE targets robust policy deployment
RoboBRIDGE proposes modular interfaces that connect learned policies with robust real-world robot execution.
Robots exchange task meaning instead of raw data
A collaborative-robot demo explores semantic communications that transmit task-relevant meaning rather than full raw data streams.
Speech-conditioned grasping reduces interaction latency
Speech2Grasp adapts text-conditioned grasp detection directly to spoken commands and reports lower latency than cascaded speech recognition.
Pegasus converts human video into robot-learnable experience
Pegasus translates human demonstrations into robot-conditioned data using task graphs, affordances, constraints, and physics verification.
Robot morphology enters the generative-design loop
A diffusion-transformer approach jointly models robot embodiment, states, and actions to optimize hardware for target motions.
