Outreach at the Cambridge Festival
The booth drew attention from both an academic audience (interested in the mechanics of training reinforcement learning policies in simulation) and a younger audience (interested in playing and making their own video games).Abstract
In this outreach session, researchers presented interesting aspects of their research to the general public. In this session, I presented a football environment that I designed in VMAS. The purpose of the environment is to evaluate our novel reinforcement learning (RL) methods, testing their ability to produce emergent strategies in a high-level task. There are several different types of agents that can compete in the football environment—in addition to RL agents, there is a non-learning-based “game AI” that exhibits dribbling, passing, defending, and shooting behaviours. During the session, I explained the goal of our research and the utility of high-level simulation environments to groups of vistors. During the presentation, people were free to step forward to compete against each other or against the game AI in the football game.
