Humanoid Robot Competes in Real-Time Tennis

There’s a new humanoid robot that’s shaking things up in the realm of tennis. Unlike most robotic competitors, which tend to rely on scripts or remote controls, this one interacts in real time on the court. It can react instantly to shots, making for an engaging experience.

This robot, roughly 4 feet tall with a compact humanoid design, comes from Galbot Robotics. They released a video demonstrating its capabilities as it went head-to-head with a human player. The system it operates on, known as LATENT, runs on the Unitree G1 platform.

It’s not merely about hitting the ball back; the robot adjusts and competes actively, showcasing impressive adaptability during live play.

What Sets This Tennis Robot Apart?

Unlike typical athletic robots that follow pre-programmed sequences, this one reacts spontaneously to human opponents. It can track fast-moving balls, navigate the court, and return shots with remarkable precision. It adapts to varying trajectories and unpredictable strikes during rallies, thanks to quick response times and synchronized movements.

How AI Learns to Play Tennis

Teaching a robot tennis is no small feat. The sport entails:

• Ball speeds reaching 107 mph
• Split-second racket contact
• Constant movement across a large court

Collecting comprehensive data on human gameplay is challenging. So instead, researchers took a unique approach.

Using Motion Fragments for Training

Rather than capturing full matches, they isolated specific movement segments:

• Forehand
• Backhand
• Side steps

They gathered about five hours of data from five players practicing on a compact 10 x 16-foot court, which is significantly smaller than a standard tennis court but still offers a valuable training ground.

How the Robot Developed Its Skills

The system learns individual movements before combining them into a cohesive sequence. This design allows the robot to:

• Approach the ball
• Control hits and make strategic plays
• Recover and reposition after each shot

To refine its capabilities further, the team trained the model within simulated conditions, adjusting variables like mass and friction. This helps the robot handle the unpredictability of live games.

Performance Against Humans

During tests, the system achieved a remarkable 96% success rate on forehand shots in simulations. In actual gameplay, it can consistently rally with human players, making accurate returns.

Watching a demonstration, the robot appears competitive. It occasionally places shots deliberately to outmaneuver its human opponent, hinting at a nascent level of decision-making. However, it does have its limitations; at times, its movements are less stable than those of a trained player, particularly when facing unpredictable shots.

Broader Implications

This advancement goes beyond just tennis; it illustrates how robots can acquire intricate human skills without flawless data. The methodology could extend to various fields like:

• Football
• Badminton
• Manufacturing
• Search and rescue operations

It’s noteworthy that this technique may greatly benefit tasks lacking complete motion data.

Future of Robotic Competition

The pathway ahead is likely clearer. As robots continue developing, it’s anticipated they will eventually train alongside and compete against professional athletes. We could be looking at exhibition matches between humans and machines in the not-so-distant future.

Conclusion

This demonstration showcases just how rapidly advancements are taking place. Robots are evolving from scripted actions to dynamic responses in real-world situations. It’s intriguing to think about whether you’d still want to compete or practice with a robot that might outperform you on the court. What are your thoughts on this?