AI Pillar Hub

Real-WorldSANDOresearch

SANDO research targets safe robot flight through unfamiliar environments

Researchers have developed SANDO, a planning method designed to keep a robot’s flight path free of collisions. The method could help flying robots navigate dangerous environments they have not previously mapped.

Source · MIT News Robotics

Collision-safe planning could improve the usefulness of aerial robots in disaster response.

All robotics storiesAgent stories: research
Real-World

ESCAPE · research

ESCAPE explores egocentric sensing for robot collision avoidance

Researchers introduce ESCAPE, an online framework for egocentric sensing, collision avoidance, and policy execution. The work addresses blind spots and sensitivity to occlusion, clutter, and calibration associated with external-camera visuomotor policies.

Source · arXiv cs.RO

The framework targets perception limitations that complicate robot operation in unstructured environments.

Real-World

Humanoid walking control research · research

Humanoid control study tackles walking on unpredictably moving ground

A new research paper addresses humanoid walking when the supporting ground moves in unknown ways. It formulates a control problem involving persistent disturbances and the changing dynamics of legged motion.

Source · arXiv cs.RO

Accounting for moving ground addresses a practical obstacle to using humanoids in unstable environments.