Interacting seamlessly with the physical world is still a big challenge for AI. Unknown properties of the environment combined with a continuous action space means that the outcome of physical actions cannot be accurately predicted. We use physical interaction games such as Angry Birds to develop AI capable of interacting with the physical world.
As part of this project, students can experiment with different AI methods and analyse which ones are most successful and which ones bring AI closer to human performance. The goal of the project is to develop a game playing AI agent that is competitive with the current state of the art of the chosen game. Students may have the opportunity to participate in an international AI game competition.
- Experience with different AI methods, such as machine learning, reasoning, or heuristics is beneficial.
- Experience with game development engines such as Unity is beneficial.
- Generally, good programming skills and good problem solving skills are beneficial.