MPC-guided Imitation Learning of Neural Network Policies for the Artificial Pancreas
This work addresses the challenge of resource-constrained medical devices by providing a safer and more efficient method for insulin therapy, though it is incremental as it builds on existing MPC and imitation learning techniques.
The paper tackles the problem of insulin control in artificial pancreas devices by using imitation learning to train neural network policies from MPC demonstrations, resulting in computationally efficient policies that outperform traditional MPC with state estimation and generalize across patient cohorts.
Even though model predictive control (MPC) is currently the main algorithm for insulin control in the artificial pancreas (AP), it usually requires complex online optimizations, which are infeasible for resource-constrained medical devices. MPC also typically relies on state estimation, an error-prone process. In this paper, we introduce a novel approach to AP control that uses Imitation Learning to synthesize neural-network insulin policies from MPC-computed demonstrations. Such policies are computationally efficient and, by instrumenting MPC at training time with full state information, they can directly map measurements into optimal therapy decisions, thus bypassing state estimation. We apply Bayesian inference via Monte Carlo Dropout to learn policies, which allows us to quantify prediction uncertainty and thereby derive safer therapy decisions. We show that our control policies trained under a specific patient model readily generalize (in terms of model parameters and disturbance distributions) to patient cohorts, consistently outperforming traditional MPC with state estimation.