This is my work in progress project making a deep q-learning agent for the Battlesnake competition. The agent currently can't compete with other agents, nor are there obstacles, it can only try and survive and eat fruit to grow.
Todo list:
- Overhaul how we save models to restart training from last session
- Experiment with different models
- Add in fruit observations to agent's 'vision', then train
- Add walls to the game and to the agent's vision
- Add simple enemy agents
- Train against other deep q-learning agents