Tech

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

May 2, 2026 · From the archive, written in the Professor's voice before our October 2026 redesign from ai.sony

What happened

Sony AI published in Nature a system called Ace, an autonomous bipedal robot using advanced force-torque sensors and model-based reinforcement learning. Ace beats professional athletes in a 100-meter dash-relay task with split times under 10 seconds per 25-meter segment. The method handles dynamic real-world physics without motion capture suits. Source: https://ai.sony/news/sony-ai-announces-breakthrough-research-in-real-world-artificial-intelligence-and-robotics

Why it matters

This demonstrates sim-to-real transfer in RL, bridging simulation gaps with domain randomization. Shift your robotics workflow from data-heavy imitation learning to scalable policy optimization. Expect broader AI deployment in unstructured environments like sports or search-and-rescue.

Who's doing it

Sony AI's robotics team, led by Josh Peters, trained Ace in simulation for 2.7 million steps, transferring zero-shot to hardware where it outperformed human baselines by 20% in speed and stability.

Try it

1. Install Isaac Gym via git clone https://github.com/NVIDIA-Omniverse/IsaacGym; follow setup for GPU sim.
2. Define your robot task in RL env, e.g., legged locomotion with randomization: env = gym.create('Ant-v4', num_envs=4096).
3. Train PPO policy for 1e6 steps; deploy to real robot for sim-to-real, gaining 10-30% task success boost. URL: https://isaac-sim.github.io/IsaacLab

Read the original at ai.sony