All experiments

Multi-agent reinforcement learning

Hide and Seek

Watch two pretrained agents use cover and line of sight. Change the arena and compare their learned behavior with the initial policies.

JavaScript · Three.js · PPO

View source on GitHub

Try it

All computation stays in your browser.

Two agents navigating a three-dimensional arena with walls and movable cover.

Opening Hide and Seek…

From the experiment

Captured from the working tool.

Two pretrained PPO policies play the same round. Walls interrupt visibility; each agent acts from its own observation.
Inspect the seeker's visible area, then rearrange the cover to see how the learned policies respond.

Source and documentation

The repository includes the implementation, reproducible examples, and tests.

Explore Hide and Seek on GitHub