The UiT Machine Learning Group is heading a interdisciplinary project that aims at developing control algorithms for automated type 1 diabetes management. The goal of our research is to use ideas from artificial intelligence to improve the daily lives of people that are living with type 1 diabetes. Interested readers can read more about the project here.

Type 1 diabetes is characterized by the lack of insulin-producing beta cells in the pancreas. The artificial pancreas promises to alleviate the burdens of self-management. While the physical components of the system – the continuous glucose monitor and insulin pump – have experienced rapid advances, a technological bottleneck remains in the control algorithm, which is responsible for translating data from the former into instructions for the latter.

At the heart of state-of-the-art artificial intelligence research lies Reinforcement Learning — the study of how to learn from unknown environments. RL is particularly suited in situations where:

• Decisions are made sequentially along a timeline.
• The actions taken depend on an observed state that is changing over time.
• The effects manifest at later points in time than the actions that induced them, and there is some notion of preferred state(s).

All of these features are certainly present in the type 1 diabetes controller challenge