​

4 min presentation in DD2438

Group 4:15

Pontus and Antonin

https://docs.google.com/presentation/d/10fdgHzWh4VqBRYdcmHABIQKUjNjXZAaTwHMflRDPL7M/edit?usp=sharing

This week:

• Get familiar with new environment, set up everything and get the project working
• Testing out ideas with Reinforcement Learning: Battlesnake Gym (https://github.com/ArthurFirmino/gym-battlesnake)

​

Overview of our plan so far...

• Use Reinforcement Learning (Q-learning)
• Goal : Learn optimal policy for the snake, avoid walls and the other snake, trap the other snake if possible
• Q gives each action for each state a value that has to be learnt
• Battlesnake Gym : open source environment to train a RL model conforming to the OpenAI Gym interface, specifically made for Battlesnake

(image from DD2380 slides)

Battlesnake Gym demo

• Trains model in a separate environment that can then be used in Battlesnake

Train the RL-model against other models

• Looking into the possibility of training the model against a known good model.
• Can improve performance.
• Might take too much time.

Progress Status Week 4

• Comment to customer paying 250 000kr for the report:
• “We have a good idea of how to solve the problem. We have made some successful preliminary tests.”
• Planned Time spent: 30%
• (Out of the combined 200h)
• Actual Time spent: 15%
• Out of the combined 200h - less availability this week �due to exams
• Actual Progress: 15%
• (estimate progress towards completing assignment)
• Risk of not completing assignment: 3%

​