A Chinese humanoid robot has demonstrated real-time tennis rallying, marking a step forward in embodied artificial intelligence.

The system, developed by Galbot and Tsinghua University in Beijing’s Haidian district, can track incoming balls, predict trajectories, reposition itself, and return shots using autonomous learning and full-body coordination.

Efficient skill acquisition

The tennis robot is powered by a training framework called LATENT, developed by a team from Tsinghua University, working with Chinese AI robotics firm Galbot, that enables humanoid robots to acquire complex sports skills from imperfect human motion data.

The system focuses on decomposing tennis into fundamental motion primitives—such as forehand and backhand strokes, lateral shuffles, and crossover steps—allowing robots to learn in a structured and scalable manner.

Rather than relying on high-precision motion capture or detailed kinematic modeling, LATENT uses “quasi-realistic” inputs derived from amateur players. Approximately 5 hours of motion data were collected using a compact capture setup, yielding noisy yet meaningful representations of human movement. This data is mapped into a latent action space, where the robot can interpret, refine, and recombine motion elements into coherent actions.

The framework integrates reinforcement learning with large-scale simulation, enabling the system to adapt to varying ball trajectories and gameplay conditions while preserving fluid, human-like motion. The trained policy was deployed on the Unitree G1 humanoid robot, where it demonstrated reliable ball striking and controlled returns.

The team claims that by reducing dependence on high-quality datasets and enabling learning from imperfect inputs, LATENT addresses a key bottleneck in robotics: replicating fast, dynamic, and precise human behaviors in unstructured environments.

High accuracy returns

The team claims that the humanoid robot is capable of tracking incoming balls, predicting their trajectory, adjusting its position, and returning shots using autonomous learning, whole-body coordination, and real-time decision-making. It can sustain multi-shot rallies against players of varying ages and skill levels, demonstrating adaptability in dynamic gameplay scenarios.

Its forehand return success rate reached 90.9 percent, marking progress in one of the most challenging benchmarks for humanoid robots—operating effectively in fast, adversarial, and unstructured environments, reports China Daily.

In earlier evaluations, researchers tested the LATENT system in real-world matches, where humanoid robots competed against human players across both forecourt and backcourt regions. Across 10,000 trials, the system delivered strong performance in both forehand and backhand strokes, surpassing previous methods in accuracy, consistency, and motion naturalness.

According to the team, at peak performance, the robot achieved a 96.5 percent success rate, consistently returning balls close to intended target areas. While it does not yet match the speed and precision of professional players, the system demonstrated the ability to maintain extended rallies and adjust to different playing styles and conditions, highlighting steady progress toward more capable, real-world athletic robots.

Recommended Articles