Public projects Pacman con IA (CS188, Berkeley)
AI · Software
Pacman con IA (CS188, Berkeley)
A Pacman agent trained with reinforcement learning, as the final project of Berkeley's online CS188 Artificial Intelligence course.
Final project of the online CS188.1x «Artificial Intelligence» course from the University of Berkeley. The agent learns to play Pacman on its own through reinforcement learning.
The agent's policy is computed with the Bellman equations: it estimates the value of each state and action from the rewards, and improves its strategy through trial and error. It's the same family of estimation and control techniques that appears in the robotics and drone projects.
