Posts

Showing posts with the label #reinforcement learning

Blue - Berkeley's Human Safe Robot

Image
Safe Robotic Co-Worker Source:  UC Berkeley's Blue Learns on the Job with Humans The roboticists at the University of California Berkeley call it Project Blue.  Blue robot is AI enabled and can safety learn tasks while working side by side with humans. It learns by reinforcement learning or trial and error. The Berkeley team has created what they call a "human safe" robot. Weaker Robot is a Better Robot The team worked on the premise that the stronger the robot the more dangerous it can be to humans.  When a robot makes a mistake, it can damage itself and people nearby.  As a result, Blue features flexible plastic arms with rounded edges that can be easily pushed away.  The team purposefully designed a weak robot that can perform significant, useful tasks because "weaker is safer". Low Cost The team says Blue is a low-cost, backdrivable, 7-DOF robotic manipulator.  The robot not only safely works side by side with humans.  It can also make...

Important Innovations Collection: Robotic Gliders Flying Like Birds

Image
Robotics Gliders Soaring Like Birds Source:  Woods Hole Oceanographic Institution Photo Montage University of California Scientists Train Autonomous Gliders to Fly Like an Eagle This is very important research that may have ramifications in terms of navigational systems and artificial intelligence for unmanned aerial vehicles (UAV's).  Physicists at the University of California San Diego have taught robotic gliders to soar like birds.  Using reinforcement learning, the gliders ride rising currents of warm air to soar just like birds.  For a great news blog, go to Important Innovations Collection: Robotic Gliders Flying Like Birds : Physicists Train Autonomous Gliders to Soar Like Eagles Source:  Woods Hole Oceanographic Institution Photo Montage New Navigational ...