'SwingBot' Learns How to Stably Handle and Swing Objects
Swinging an object is challenging for robots because it requires knowledge about the object's center of mass and friction. "SwingBot" from MIT can learn these properties using tactile gestures like shaking and tilting. The robot doesn't use cameras or vision methods; instead, its grippers use GelSight tactile sensors that measure the pose and force distribution of the object. SwingBot is able to predict the swing angle achieved by a robot performing dynamic swing-up manipulations on a previously unseen object.
More From SAE Media Group
Motion Design
Gripper Handles Freely Moving Cables
Tech Briefs
Algorithm Gives Robots a Faster Grasp
Motion Design INSIDER
Machine Intelligence to Build Soft Machines
Robotics & Automation INSIDER
Autonomous Aerial Robots Prioritize Rooms in Multiroom Exploration
Tech Briefs
Dexterous Robotic Gripper Can Manipulate Cables
Motion Design
Taking Industrial Robot Automation to New Heights
Tech Briefs
Robot Reads Braille Twice as Fast as Humans
Tech Briefs
Giving Robots the Human Touch via Listening
Motion Design
Building More Dexterous Robots
Tech Briefs
Soft Robotics Perception System
Robotics & Automation INSIDER
AI Yields ‘Athletically Intelligent’ Robotic Dog
Motion Design
Teaching Robots the Craft of Manipulating Materials
Motion Design
Autonomous Robot Opens Doors
Tech Briefs
A Robotic 'Digger Finger' Senses Buried Items
Tech Briefs
HERMES: Two-Legged Robot with Human Reflexes
Tech Briefs
Robots Get A Feel For The World
Tech Briefs
Algorithm Designs Soft Robots That Sense
Transcript
00:00:00 [Music Playing]