Applying Bayesian Optimization to Brain Injury Rehabilitation Games
After a brain injury, patients suffer a reduction on their cognitive and motor skills. Their upper body dexerity is reduced, or their visuo-spatial ability is hindered. In all these cases, rehabilitation that leverages brain plasticity has the potential of helping the patient recover these skills. Unfortunately, the repetitive nature of these exercises makes them boring.
Rehabilitation Games, a subset of serious games, have the potential keeping the patient engaged if the level of difficulty adapts to the skill of the player, which varies from patient to patient, and even from session to session.
In this project, we want to automatize the process of finding the ideal difficulty in these Rehabilitation Games by applying some recent developments in adaptive robotics: the [Intelligent Trial-and-Error algorithm](https://youtu.be/T-c17RKh3uE). We start by building several levels using Procedural Content Generation, we assess their difficulty using automated bots, and we model and update the difficulty of these levels using Gaussian Processes.
This project has confirmed collaboration with the [Center for Rehabilitation of Brain Injury](https://cfh.ku.dk/english/) in Copenhagen, and has received funding from the Google Faculty Award 2019. If you are interested in knowing more about this project, feel free to contact us at migd(at)itu(dot)dk.
Research outputs:
– [Finding Game Levels with the Right Difficulty in a Few Trials through Intelligent Trial-and-Error](https://arxiv.org/abs/2005.07677)