Skip to content
Saketh Kamatham

Super Mario Bros via Deep Q-Learning

A course project before my thesis: Q-learning and Double DQN agents that learned to complete a Super Mario Bros level autonomously.

Status
Archived
Tags
  • Reinforcement Learning
  • DQN

Course project, before the thesis.

Problem

Train an agent to autonomously complete a Super Mario Bros level. An earlier, smaller reinforcement-learning project that came before the thesis and informed how I approached it.

Approach

Implemented Q-learning and Double DQN (DDQN) agents against the game environment.

Outcome

Both agents learned to complete the level autonomously.