Training humanoid robots through trial and error is proving to be more effective than the traditional engineering approach. Instead of manually programming a robot’s movements, researchers are now utilizing reinforcement learning techniques, allowing robots to learn tasks independently. This shift marks a significant advancement in robotics, eliminating the need for complex programming and relying on the robot’s ability to figure out tasks by itself.
Recent developments have showcased the capabilities of humanoid robots in various activities, such as playing tennis. By using imperfect human motion capture data, robots can learn essential skills like forehand, backhand, and footwork. These motion fragments are then pieced together using reinforcement learning algorithms to enable the robot to perform complex tasks like playing tennis efficiently.
The latest achievement in this field is a humanoid robot that can play tennis by returning the ball and sustaining a rally. Although the robot has no game strategy and does not aim to score points, its ability to perform these skills is impressive and somewhat eerie. Referred to as LATENT (Learns Athletic humanoid TEnnis skills from imperfect human motioN daTa), this system combines various motion fragments to train the robot in playing tennis strokes accurately.
The training process takes place in a simulated environment before being transferred to a real-world humanoid robot. The research paper highlights the success of this method by stating that the robot can sustain multi-shot rallies with human players efficiently. This breakthrough signifies a significant milestone in humanoid robotics, demonstrating the potential for robots to learn and perform complex tasks with precision.
As we look back on the evolution of humanoid robots, from attempting to walk like humans to mastering sports activities like tennis, it is evident that the future holds endless possibilities for robotic advancements. Stay updated with the latest articles on robotics and technology by subscribing to our newsletter, RSS feed, and following us on social media platforms.
About the Author
