Advances in continuous glucose monitoring (CGM) technologies and wearable devices are enabling the enhancement of automated insulin delivery systems (AIDs) towards fully automated closed-loop systems, aiming to achieve secure, personalized, and optimal blood glucose concentration (BGC) management for individuals with diabetes. While model predictive control provides a flexible framework for developing AIDs control algorithms, models that capture inter- and intra-patient variability and perturbation uncertainty are needed for accurate and effective regulation of BGC. Advances in artificial intelligence present new opportunities for developing data-driven, fully closed-loop AIDs. Among them, deep reinforcement learning (DRL) has attracted much attention due to its potential resistance to perturbations. To this end, this paper conducts a literature review on DRL-based BGC control algorithms for AIDs. First, this paper systematically analyzes the benefits of utilizing DRL algorithms in AIDs. Then, a comprehensive review of various DRL techniques and extensions that have been proposed to address challenges arising from their integration with AIDs, including considerations related to low sample availability, personalization, and security are discussed. Additionally, the paper provides an application-oriented investigation of DRL-based AIDs control algorithms, emphasizing significant challenges in practical implementations. Finally, the paper discusses solutions to relevant BGC control problems, outlines prospects for practical applications, and suggests future research directions.

Objective: The interpretation of time series data collected in free-living has gained importance in chronic disease management. Some data are collected objectively from sensors and some are estimated and entered by the individual. In type 1 diabetes (T1D), blood glucose concentration (BGC) data measured by continuous glucose monitoring (CGM) systems and insulin doses administered can be used to detect the occurrences of meals and physical activities and generate the personal daily living patterns for use in automated insulin delivery (AID). Methods: Two challenges in time-series data collected in daily living are addressed: data quality improvement and the detection of unannounced disturbances of BGC. CGM data have missing values for varying periods of time and outliers. People may neglect reporting their meal and physical activity information. In this work, novel methods for preprocessing real-world data collected from people with T1D and the detection of meal and exercise events are presented. Four recurrent neural network (RNN) models are investigated to detect the occurrences of meals and physical activities disjointly or concurrently. Results: RNNs with long short-term memory (LSTM) with 1D convolution layers and bidirectional LSTM with 1D convolution layers have average accuracy scores of 92.32% and 92.29%, and outperform other RNN models. The F1 scores for each individual range from 96.06% to 91.41% for these two RNNs. Conclusions: RNNs with LSTM and 1D convolution layers and bidirectional LSTM with 1D convolution layers provide accurate personalized information about the daily routines of individuals. Significance: Capturing daily behavior patterns enables more accurate future BGC predictions in AID systems and improves BGC regulation.

Traditional projection dynamic monitoring methods focus on simultaneously decoupling the correlations among the process variables and the autocorrelations of the variables, which leads to a mixing problem of the two correlations. The mixing problem decreases the ability to model the dynamic correlations, thus decreasing the detection rates (DRs) of some faults. Considering that the projection matrix may cause the mixing of the two correlations, this paper proposes a dynamic monitoring method based on directly monitoring original variables. This paper firstly adopts the kernel least mean squares (KLMS) method to establish a univariate dynamic model, then adopts the univariate dynamic model and the Dual Control Chart (DCC) to build the multivariate direct monitoring method, which is named the KL2C method (KL represents KLMS. and 2 C represents DCC). Then, the Dynamic Nearest Neighbor Kernel Space (DNNKS) is proposed to overcome the redundant dimensions problem of the KL2C method, which is named the DKL2C method (D represents DNNKS). Furthermore, based on the Joint Control Chart (JCC), this paper puts forward a dynamic monitoring method of two sequences, which both considers the advantages of the projection dynamic monitoring method and the direct dynamic monitoring method together. Finally, this paper utilizes the Tennessee Eastman (TE) process and the continuously stirred tank reactor (CSTR) process to verify the effectiveness of the proposed methods.

The artificial pancreas (AP) systems based on model predictive control (MPC) are expected to provide effective and safe regulation of blood glucose concentration (BGC) for people with type 1 diabetes. However, the development of AP systems based on MPC is challenged by the high computational burden of MPC facing limited computational resources and energy supply of portable devices. In this article, we proposed a novel event-triggered MPC (ET-MPC) to reduce the computational burden and lower the energy consumption of fully automated AP systems. The predictor in the ET-MPC is based on a hybrid glucose prediction model that incorporates prior knowledge of glucose dynamics. By exploring the distribution of modeling residuals, it enables the model to capture changes in glucose concentration dynamics and avoid unnecessary updates of model parameters. Event-triggered (ET) rules aim to lower the energy consumption of the optimizer, realized by reducing the times of solving optimization problems. They are proposed by leveraging the physiological insights of glucose dynamics and fundamental mechanisms of MPC. The efficiency and advantages of the proposed ET-MPC are illustrated through preclinical simulation studies challenged by unannounced meals and physical activity. Overall, the predictor and optimizer are updated for 2.4% and 28.8%, respectively, of time over the closed-loop experiment period with the appeal that BGC is maintained within the safe range for 79.7% of time, which is not significantly different from the results when ET rules are not utilized. The proposed ET-MPC enables the AP system to manage BGC effectively with a considerably low computational burden and energy consumption.