Robots can now train themselves with new "practice makes perfect" algorithm

– Robots

MIT and The AI Institute developed the "Estimate, Extrapolate, and Situate" (EES) algorithm, enabling robots to identify skill weaknesses and practice autonomously for improvement.

– New Algorithm

The algorithm was tested on Boston Dynamics' Spot robot, significantly reducing practice time for tasks like sweeping and handling objects, enhancing efficiency.

– Robot Testing

Download App

Researchers envision this technology improving robots in various environments, such as factories and households, with potential for combining virtual and physical practice.

– Future Applications

Download App

Future development aims to create algorithms that can reason over sequences of practice, helping robots learn complex tasks independently, essential for widespread home use.

– Long-term Goals

Download App