Sony Ace robot beats pros at table tennis
Sony AI released Ace, a seven-degree-of-freedom arm equipped with event-based cameras and trained via reinforcement learning on 10 million ball trajectories. In official matches Ace won 52 percent of points against a top-50 Japanese professional. The system updates its policy every 50 milliseconds using on-robot GPU inference.
You see how simulation-to-real transfer works when the reward function matches real physics. This changes your thinking from pure software prototyping to closing the loop with physical sensors and rapid policy updates.
Sony AI published the full results in Nature Robotics; their internal league table shows Ace maintaining a 0.71 win rate across 300 matches against regional pros in Tokyo.
Step 1: Clone the Ace research repository at github.com/SonyAI/Ace-RL and install the provided Docker environment. Step 2: Launch the included MuJoCo table-tennis simulation and train a policy for 500k steps using the supplied SAC algorithm. Step 3: Transfer the trained weights to the open-source robot arm model and record success rate on 100 simulated rallies.