SMU Data Science Review

Rocket Learn

Daanesh Ibrahim, Southern Methodist UniversityFollow
Jules Stacy, Southern Methodist UniversityFollow
David Stroud, Southern Methodist UniversityFollow
Yusi Zhang, Southern Methodist UniversityFollow

Abstract

Abstract. This paper covers the development, testing, and implementation of Reinforcement Learning methods designed to autonomously learn and optimize Rocket League play. This study aims to analyze and benchmark model frameworks commonly used in Reinforcement Learning applications. These models can be applied to tasks ranging in difficulty from simple to superhumanly complex, and this study will begin with and build upon simple models performing simple tasks. It will result in complex models performing difficult tasks. Models will be allowed to train autonomously on the game using mass parallelization to expedite training times with the goal of maximizing reward function scores. This research constructs a framework to identify the best performing Reinforcement Learning model to complete this task. Multiple Reinforcement Learning methods were attempted and it was found that a Proximal Policy Optimization (PPO) model was able to learn how to play the game and consistently increase its reward function scores over time. Of all the models attempted for this game, PPO did the best job of learning how to play and it is recommended for future tasks in similar spaces.

Recommended Citation

Ibrahim, Daanesh; Stacy, Jules; Stroud, David; and Zhang, Yusi (2021) "Rocket Learn," SMU Data Science Review: Vol. 5: No. 2, Article 12.
Available at: https://scholar.smu.edu/datasciencereview/vol5/iss2/12