Project Background
Study Program
MSc. Information Engineering and Computer Science
Hochschule Rhein-Waal
Semester
Summer Semester 2017
Course
Applied Research Project A (Group Project)
Instructors
Prof. Dr. Timo Kahl
Introduction
Deep Learning is currently at the vanguard of recent AI research. It enables applications to learn solving varying tasks, like colorizing gray-scale imagery, improving the energy efficiency of data centers, increasing the performance of CPUs and to play board and video games. Especially video games (or game simulations in general) lead to recent breakthroughs in Deep Learning. The most impact had DeepMind’s AI, which is capable of playing many Atari 2600 games.
The game Beastly Rivals Onslaught (BRO) is an experimental subject for research in artificial intelligence (AI) developed in a previous project. This scope of this project is to examine if BRO and its AI framework are a suitable use-case for developing and implementing learning AI.
Main Responsibility
- research on Nature Inspired Algorithms and Deep Reinforcement Learning
- research on related work contributions
- implementation of AI’s output actions
- pass the ball to the enemy which is in the sight of the ball carrier, i.e. demanding the least rotation effort
- pass the ball to the enemy with the highest score
- pass the ball to the enemy with the lowest remaining lives
- pass the ball to the enemy who is in the moving direction of the beast
- double checking DQN implementation
- assist in development of teaching concept based on this project