Putting some of the best local models to the development test ...
Two reinforcement-learning agents for Snake on a 20 by 20 grid, measuring how much of the score is the network versus hand-coded search. An autonomous agent trained via Double Deep Q-Networks (DDQN) ...
🛞 > what do you see? → rover_see("front") [inline image → model SEES it] 🤖 I'm on a sidewalk. Tree ahead-left, clear path forward. 🛞 > drive up to the tree, carefully → rover_navigate(steps=[...], ...