The Neuron Daily reported on Skild AI's simulation-trained soccer policy and its transfer to a physical humanoid robot.
ConcludedRan 3 days2 sourcesupdated 25 Sep 2026
Skild AI Trains Humanoid Robot to Play Soccer via 140 Years of Sim Self-Play
The gist
Transferring a policy learned entirely in simulation to a physical humanoid robot for a dynamic, multi-skill task like soccer is a concrete test of sim-to-real transfer in reinforcement learning. The task requires the robot to manage several physical and strategic challenges at once.
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.
The full picture
Skild AI trained a humanoid robot called the Messinator to play soccer using 140 simulated years of self-play reinforcement learning, then transferred that learned policy to a physical robot. The robot generates joint-angle commands at 50Hz rather than selecting from a discrete action space. Training worked by having the robot repeatedly compete against saved earlier versions of its own policy across millions of accelerated virtual soccer games. The system handles balance, ball control, collision recovery, opponent modeling, locomotion, and strategy simultaneously.
How it developed
Rohan Paul published details of Skild AI's Messinator humanoid soccer robot trained via 140 years of simulated self-play.
Sources
Rohan Paul on X
Get the daily email
When stories like this move, I write about them in an email each morning.
One free email each morning, around 8 AM US Eastern.
Check your email.
I have sent you a confirmation link. One click and you are in.
Not in your inbox? Look in spam or promotions. If it's there, mark it "Not spam" so future editions land in your inbox.