BSc (Hons) Computer Science (Artificial Intelligence) is the new cutting edge degree bringing the specialism of Artificial Intelligence to the fore.
The projects currently being displayed are from our current academic year. To view other years
please select below.
Oliver Guy
End to End Lane Keeping In Euro Truck Simulator 2
The aim of this project is to demonstrate that an end to end lane keeping system can be fully developed within a virtual environment. This goal can be broken down into three core components; training pipeline, game control and the end to end model. The finished project should have two elements, Firstly a series of scripts that allow the user to parse and label collected data and train a e2e model. And secondly, a script that allows the user to demonstrate that model in the game environment.
The aim of this project is to demonstrate that an end to end lane keeping system can be fully develo...
Reef Lakin
Mastering Catan Using Multi-Agent Reinforcement Learning
Catan, known better perhaps by its previous and longer name, Settlers of Catan, has been gracing the shelves of board game fans with its presence since 1995. It's widely considered to be one of the most popular "Eurogames" of all time. For good reason: the game has a random setup each time its played, so as well as being generally a very entertaining experience, it holds a lot of replayability that many other popular board games only wish they could possess. It is this pervasive randomness that makes Catan the ideal subject for framing one of machine learning's more visually attractive sub-fields: multi-agent reinforcement learning. My work on an AI-driven Catan agent for this project, which I am creatively naming CatanAI*, adds to a catalogue of similar work by machine learning researchers who have been studying Catan in this way since as early as 2004. The goals of my project are as follows: *Create a bespoke Catan game environment from scratch, using PythonAdd the ability for a virtual agent to play CatanEnable the agent to learn from its mistakes and master Catan, in an environment where similar agents are also competing for the same thingAllow Catan fans to play against my virtual agents. The main thing I want to get from this project is learning how small changes to reinforcement learning methods can have a big impact on efficacy of agent training, myself exploring the many intricacies of RL problems and multi-agent models.I also love Catan, so any excuse to engage with it through this dissertation is very valuable.*because this was the working title of my project and it eventually just stuck.
Catan, known better perhaps by its previous and longer name, Settlers of Catan, has been gracing the...