You like video games and are fascinated by artificial intelligence? Then this is the right place for you, because the Institut für Informationsverarbeitung (TNT) started an AI Competition Team and is looking for motivated students. Here you have the chance to apply the knowledge you have learned at university. You can also earn credit points in the form of a student research project (Studienarbeit).
The aim is to develop AIs for selected games and participate in competitions with it.
Interested or still have questions? No problem, just contact us under: email@example.com
BattleSnake is based on the classic video game Snake. Each player controls a snake on a 2D grid and has to ensure that it does not collide with itself, another snake, or the borders of the game. Additionally, the snake has a finite number of health points which decay over time but can be topped up by eating food which randomly pops up. Several snakes compete against each other at once, and the last snake to survive wins.
The annual BattleSnake competition in Victoria, BC, Canada regularly welcomes hundreds of contenders. Those interested can either compete individually or in teams of up to five persons and win cash prizes up to $1000.
In addition, it is now possible to challenge other snakes from all over the globe on play.battlesnake.io.
At the 2019 competition, our students Maximilian Schier, Niclas Wüstenbecker and Frederik Schubert took second place in the intermediate category. Read more...
Battlecode is a real-time strategy game, in which two teams, each consisting of up to four members, compete against each other to defeat the AI of the competitor. The competition is organized in five sub-tournaments (Sprint Tournament, Seeding Tournament, Qualifying Tournament, Newbie und Final Tournament). In the Final Tournament, cash prizes totaling $50,000 can be won.
On the Battlecode website online lectures as well as the source codes of the winners of the past years can be found. This allows a relatively easy entry.
Since fall 2018, we additionally organize a Machine Learning for Game AIs lab at TNT. In the first part the fundamental knowledge is learned - especially in the area of path planning, machine learning and reinforcement learning. In the second part, each team develops an AI for Battlecode.
One of our Battlecode teams in 2019 consisting of Niclas Wüstenbecker, Paul Hindricks, and Maximilian Schier was very successful. They qualified for the final which took place at MIT in Boston/Cambridge and made it into the top 10.
The final 2019 can be watched again on YouTube. The student team from our lab is called Wololo.
The General Video Game AI Competition (GVG-AI) is about developing an agent who can not only play one particular game well, but all of them. The agent should be able to see for himself what game he is playing and how he has to play it.
Besides the "Single Player Learning Track", where an agent should learn to play different games like in the other competitions, there are other tracks, which cover different aspects of learning. The full list of tracks is:
Single Player Learning Track
Single Player Planning Track
2-Player Planning Track
Level Generation Track
REule Generation Track
The organizers of the Competition offer a large framework with various games for training and testing. The framework is written in Java using the Video Game Definition Language (VGDL), but there is also a Python environment for the learning track. Several tracks are usually combined into one competition at conferences dealing with video games. All you have to do is submit your own agent in time via the official website.
One agent for many games
Programming languages: Java or Python
Competitions take place at conferences around the world