The Ghostly and Ghostly+ research projects have resulted in a set of platform games
controlled by electromyography (EMG) for rehabilitation and strength training purposes. The
generation of rich levels of such 2D platform games is time consuming. AI is used more and more often
for such tasks. A variety of approaches exist for generating artificial game levels. It is however not
always easy to predict how difficult users will find these levels and hence it is also difficult to generate
levels of a known difficulty level. This is an issue in normal gaming and even more in rehabilitation
gaming, due to the heterogeneity of the cognitive and motor skills of the subjects, their motivation
etc.
Next to AI methods for generating platform games, there exist plenty of highly successful methods for
training an AI agent to play platform games. The success of such agent in the level could be a proxy
for the human perceived difficulty level.
Linking the two could enable us to artificially generate levels of know/predictable difficulty level by
seeing it as a search/optimisation problem: the first agent generates candidate levels, the second
agent evaluates the difficulty: when working together, both can reinforce each other to become better
in their respective tasks and solve the problem of generating levels with known difficulty.

Literature Review (ETOC: 2 months): Familiarize with existing literature on generation of
game levels and on agents for playing platform games. Search for frameworks to couple
both.
- Selection of the right agents, basic training in the ghostly context.
- Development of a learning framework in which both are coupled.
- Basic evaluation of generated content with human players.

Following an MSc in a field related to one or more of the following: Computer
Science, Biomedical Engineering, Applied Computer Science - Digital Health.
• Strong programming skills (Python).
• Ability to write scientific reports and communicate research results at
conferences in English.