Sigma Xi Poster Competition

Training an artificial agent to play a first person shooter game

Academic Level at Time of Presentation

Senior

Major

Engineering Physics

Minor

Computer Science

2nd Student Academic Level at Time of Presentation

Senior

2nd Student Major

Engineering Physics

List all Project Mentors & Advisor(s)

Dr. James Hereford

Presentation Format

Poster Presentation

Abstract/Description

Video games are rich and challenging domains and have been used extensively for evaluating Artificial Intelligent algorithms. In this research we use Reinforcement Learning (RL) to train an Artificial Intelligence agent to play the first person shooter game VizDoom. The agent we trained to play VizDoom used images from the game at multiple frames per second as its input and then decided an action such as move left, move right, shoot, etc. We utilized the techniques of curriculum learning and reward shaping to improve the training process and achieve better results in the game.

Spring Scholars Week 2023 Event

Sigma Xi Poster Competition

This document is currently not available here.

Share

COinS
 

Training an artificial agent to play a first person shooter game

Video games are rich and challenging domains and have been used extensively for evaluating Artificial Intelligent algorithms. In this research we use Reinforcement Learning (RL) to train an Artificial Intelligence agent to play the first person shooter game VizDoom. The agent we trained to play VizDoom used images from the game at multiple frames per second as its input and then decided an action such as move left, move right, shoot, etc. We utilized the techniques of curriculum learning and reward shaping to improve the training process and achieve better results in the game.