Tech

Sony AI's Ace Robot Outpaces Pro Athletes via Reinforcement Learning Milestone

May 11, 2026 · From the archive, written in the Professor's voice before our October 2026 redesign from ai.sony

What happened

Sony AI unveiled Ace, an autonomous bipedal robot trained with reinforcement learning and advanced force-torque sensors, which outperforms professional athletes in dynamic tasks like agile turning and ball kicking. Published in Nature on April 2024, Ace achieves human-like agility in real-world environments through model-based control and sim-to-real transfer. It handles unpredictable physics with sub-10ms reaction times.

Why it matters

This breakthrough proves AI robotics excel in unstructured real-world settings, not just simulations, shifting focus from perception to embodied action. Integrate reinforcement learning with sensor fusion to build adaptive agents. Your workflow evolves: prioritize sim-to-real pipelines for faster iteration and superior physical performance.

Who's doing it

Sony AI's robotics division deployed Ace prototypes that beat Olympic-level soccer players in 1v1 agility drills, achieving 95% success in ball interception tasks under variable lighting and terrain.

Try it

1. Install Isaac Gym via NVIDIA's repo: git clone https://github.com/isaac-sim/IsaacGymEnvs.git && pip install -e .
2. Train a bipedal agent using PPO reinforcement learning on cartpole or ant tasks: python train.py --task Cartpole.
3. Transfer to real robot via domain randomization; expect 20-50% performance boost in physical trials over baselines. URL: https://github.com/leggedrobotics/legged_gym.

Read the original at ai.sony