A vision-based control system called Neural Jacobian Fields enables soft and rigid robots to learn self-supervised motion control using only a monocular camera. The system, developed by MIT CSAIL researchers, combines 3D scene reconstruction with embodied representation and closed-loop control.
Robot, know thyself: New vision-based system teaches machines to understand their bodies Neural Jacobian Fields, developed by MIT CSAIL researchers, can learn to control any robot from a single camera, without any other sensors.
The robot can support the person’s full weight, lifting them from sitting to standing and vice versa along a natural trajectory. And the arms of the robot can catch them by rapidly inflating side airbags if they begin to fall.
The word “robot” was coined by the Czech writer Karel Čapek in a 1920 play called Rossum’s Universal Robots, and is derived from the Czech robota, meaning “drudgery” or “servitude”.
SPROUT is a flexible robot built by MIT Lincoln Laboratory and Notre Dame researchers to assist in disaster response. Emergency responders can use the robot to navigate and map areas under rubble to plan rescue operations.
New insect-scale microrobots can fly more than 100 times longer than previous versions. The new bots, also significantly faster and more agile, could someday be used to pollinate fruits and vegetables.
A hopping, insect-sized robot can jump over gaps or obstacles, traverse rough, slippery, or slanted surfaces, and perform aerial acrobatic maneuvers, while using a fraction of the energy required for flying microbots.
LucidSim is an AI-powered simulator that trained a robot dog to perform parkour using generated images without any real-world data. This approach, from MIT CSAIL researchers, scales up training data, helping robots transfer skills to the real world without additional fine-tuning.
MIT researchers developed a powerful system that could help robots safely navigate unpredictable environments using only images captured from their onboard cameras.
A new training interface allows a robot to learn a task in several different ways. This increased training flexibility could help more people interact with and teach robots — and may also enable robots to learn a wider set of skills.
