Robots Learn Complex Skills with AI

Researchers at UC Berkeley have developed a fast and precise method for teaching robots tasks such as assembling a motherboard or an IKEA drawer. In Sergey Levine’s Robotic AI and Learning Lab, a robot successfully executed a task called 'Jenga whipping,' learning this skill in just 1.25 hours.

The robot, using a black leather whip, was able to strike a single block from a perfectly stacked tower of 39 Jenga blocks while keeping the rest of the tower intact. This AI-powered training method allows robots to learn from human demonstrations and feedback, as well as their own attempts.

Reinforcement Learning Method

The method used in the lab is called 'reinforcement learning.' A robot performs a task in the real world, receives feedback from cameras, and learns from its mistakes. The researchers released a new software suite that enables others to quickly replicate their successes.

Human Intervention

The new version of the system incorporates human intervention. Using a special mouse to control the robot, a human can correct its actions, which is then stored in the robot's memory. As the robot learns, the need for human intervention decreases.

Complex Tasks and Adaptability

The robots underwent a series of complex tasks, such as flipping an egg in a pan and assembling various components. Researchers also tested the robots' adaptability by creating situations that required quick reactions to changes.

• The robot achieved a 100% success rate in completing tasks.
• The new training system proved to be faster and more accurate than traditional methods.

According to the researchers, such robots could be beneficial in manufacturing processes that require high precision and adaptability.