Observation Space

The glucoregulatory system is a complex non-linear system. In the glucose control problem the true system state information is not available. Instead we are limited to sensor and actuator measurements. Hence, we use an observation-space insteads of the true state-space.

In GluCoEnv, two types of observation-spaces are provided. The default observation-space named current, outputs the current glucose sensor measurement, while past_history provides an observation-space where glucose sensor and insulin pump measurements are augmented by a window of past historical values. The past_history is useful when formulating the problem as a Partially Observable Markov Decision Process (POMDP), while current is used in many open-loop clinical treatment algorithms and classical algorithms like PID and MPC.

The observation-space is a setting which should be provided when initialising the environment.