prisoners_in_R

This project is for an algorithm to play in the TU Delft EPA1315 Prisoner's Dilemma tournament.

Functionality

1. Uses an elementary genetic algorithm (based on Axelrod's method [1]) to learn the best ways to react to another agent based on their previous choices
2. Creates an agent that will be able to play with other agents in a tournament

This agent does not evolve during a game – it must be trained beforehand on some set of data. In this case, tournament.csv is from the EPA1315 Fall 2017 tournament. Since this agent uses Axelrod's method, the first three encounters will be played with a cooperate and then modified Tit-for-tat methodology. Only the n > 3 encounter will actually implement Axelrod's method.

See prisoners_in_R.Rmd for more detailed documentation.

Usage

1. Run main.py and copy the output string
2. Put the output string into Agent_BRUTE_MELLOW.R, which is the final agent we submitted.
3. Set your working directory to Example_Tournament
4. Run the example tournament *.Rmd file and see your results!

Acknowledgements

This project was made by Siemon Keij, Phillip Seijger, Mees Hoff, and Jason R Wang.

[1] S. Mittal and K. Deb, “Optimal Strategies of the Iterated Prisoner’s Dilemma Problem for Multiple Conflicting Objectives,” IEEE Trans. Evol. Comput., vol. 13, no. 3, pp. 554–565, Jun. 2009.