IFAC Journal of Systems and Control最新文献

Pharmacometric modeling plays an important role in drug development and personalized medicine. Pharmacometric covariate models can be used to describe the relationships between patient characteristics (such as age and weight) and pharmacokinetic (PK) parameters. Traditionally, the functional structure of these relationships are obtained manually. This is a time-consuming task, and consequently limits the search space of covariate relationships. The use of data-driven machine learning (ML) in pharmacometrics has the potential to automate the search for adequate model structures, which can speed up the modeling process and enable the evaluation of a wider range of model candidates. Even with moderately sized data sets, ML approaches require millions of simulations of pharmacokinetic (PK) models, which dictates the need for an efficient simulator. In this paper, we demonstrate how to automate covariate modeling using neural networks (NNs), that are trained using efficient PK simulation techniques. We apply the methodology to a propofol data set with 1031 individuals and compare the results to previously published covariate models for propofol. We use the NN as a function approximator that relates covariates to the parameters of a three-compartment PK model, and train it on dose and plasma concentration time series. Our study demonstrates that NN-based covariate modeling allows for automation of the otherwise time-consuming task of identifying which of available covariates to include in the model, and what functional mappings from these covariates to PK model parameters to consider in the model search. Additional to this saving in modeler effort, the NN-based model obtained in our clinical data set example has PK parameters within a clinically reasonable range, and slightly enhanced predictive precision than a previously published state-of-the-art covariate models for propofol model.

Recent advancements in AI have significantly enhanced smart diagnostic methods, bringing us closer to achieving end-to-end diagnosis. Ultrasound image segmentation plays a crucial role in this diagnostic process. An accurate and robust segmentation model accelerates the process and reduces the burden of sonographers. In contrast to previous research, we consider two inherent features of ultrasound images: (1) different organs and tissues vary in spatial sizes, and (2) the anatomical structures inside the human body form a relatively constant spatial relationship. Based on those two ideas, we proposed two segmentation models combining multi-scale convolution neural network backbones and a spatial context feature extractor. We discuss two backbone structures to extract anatomical structures of different scales: the Feature Pyramid Network (FPN) backbone and the Trident Network backbone. Moreover, we show how Spatial Recurrent Neural Network (SRNN) is implemented to extract the spatial context features in abdominal ultrasound images. Our proposed model has achieved dice coefficient score of 0.919 and 0.931, respectively.

This study aims to improve the accuracy of Electrical Impedance Tomography (EIT) measurements for monitoring ventilation and cardiac signal in medical imaging by proposing a new signal separation approach that does not require contrast agents. Conventionally, contrast agents like high-conductive saline solutions are used for signal separation in EIT measurements. This study uses a harmonic analysis on EIT raw voltage data to separate the ventilation- and cardiac-related signals (early separation). It evaluates its efficacy with a simulation model at low (1%) and high (10%) superimposed additive noise levels against the already published harmonic analysis at pixel level after EIT image reconstruction (late separation). The findings indicate that the voltage-based harmonic analysis approach, i.e., early separation, provides reliable signal separation, especially under high noise conditions, compared to the late separation. This method enables the possibility of incorporating independent cardiac-specific or ventilation-specific prior knowledge into the image reconstruction process, potentially improving the resulting images.

This paper investigates the use of a recently developed pharmacokinetic/pharmacodynamic model for the design of a Proportional–Integral–Derivative controller for total intravenous anesthesia. In particular, we consider the administration of propofol as manipulated variable and the BIS signal as the process variable, and we propose a personalized approach to tune the controller by using the Eleveld model. Simulation results show that the personalized controller outperforms the population-based one, which fails to provide the required clinical performance for elderly people. Thus, the development of a pharmacokinetic/pharmacodynamic model specifically devised for control design purposes would be beneficial to provide a truly personalized control law and to increase the overall performance.

Ultra-early detection of diseases with High-Dimension Low-Sample-Size (HDLSS) data has been effectively addressed by the Dynamical Network Biomarkers (DNBs) theory. After ultra-early detection, it is crucial to consider ultra-early medical treatment for the detected disease. From the viewpoint of control engineering, ultra-early medical treatment is achieved by increasing the system’s stability and preventing the bifurcation, called re-stabilization. To implement effective re-stabilization, the system matrix is necessary. However, the available data in biological systems are often HDLSS, which is insufficient to identify the system matrix. In this paper, to realize HDLSS-based ultra-early medical treatment, we investigate an HDLSS data-based system matrix estimation method. First, HDLSS data is applied to compute the sample covariance matrix of the steady state. By assuming that the system matrix is sparse and the structure of the system matrix is known, it can utilize the Lyapunov equation to estimate the system matrix from the covariance matrix. The Lyapunov equation-based method gives a unique optimal estimation if the covariance matrix is full-rank. Otherwise, the optimal estimation is not unique. The sample covariance matrix computed from the HDLSS data is not full-rank. Thus, we apply shrinkage estimation to overcome the under-determined issue to obtain a well-conditioned covariance matrix with full rank. In addition, we confirm the effectiveness of the proposed method through numerical simulations.

As the light reflex of the pupil relies on covert attention, it has been suggested for use as a non-contact, calibration-free human–computer interface. A vital aspect of the widespread adoption of this interface is the elevated information transfer rate. This study involved the design and assessment of binary light stimulation under conditions where the stimulation and discrimination systems operate asynchronously. Binary light stimulation, characterized by uniquely timed light flickering, enables the discrimination system to identify the gazed target through variations in pupil diameter. To enhance the temporal efficiency, a cyclic code was selected for binary coding. Algorithms for selecting optimal codes, determining phase-shift relationships, and designing binary codes with strategic location arrangements were developed. An experimental application of a template-matching-based classification algorithm yielded over 83% accuracy in identifying a gazed target among nine possibilities. The average information transfer rate was 30 bits/min under stable conditions. Additionally, by analyzing the values of the proposed evaluation functions, we can predict combinations prone to misclassification in the target classification. Practically, this research offers a robust method for brain–computer interfaces, potentially benefiting users with severe motor restrictions or in contexts requiring hands-free operation.

This paper proposes a novel adaptive nonlinear observer design for a voltage source converter based on high-voltage direct current (VSC-HVDC) transmission systems. We consider a system consisting of a power grid, and a converter station connected to an unknown load through a long HVDC cable. The primary contribution of this work is the development of a global high-gain observer that facilitates the estimation of all system states. Specifically, it encompasses the estimation of power grid parameters, such as the angular frequency and the voltage at the point of common coupling (PCC), as well as the states of the HVDC cable and the current absorbed by the load. The performance of the proposed observer is assessed through theoretical analysis and simulations. Additionally, we implemented our observer on a digital signal processor (DSP) eZdsp in a processor-in-the-loop (PIL) quasi-real-time setting. Experimental results, coupled with numerical simulations, showcase the outstanding performance of our proposed observer.

Non-self- regulating nature of integrating plant causes their regulation a great challenge to researchers. Presence of time delay and positive zero in such processes makes their control more demanding. A novel control technique is developed for such processes to cope with the difficult situations and achieve enhanced performance and robustness. Suggested method is twofold: first the Inner loop controller’s parameters are adjusted to stabilize the process using maximum sensitivity concept and then the outer loop control block is designed by specifying desired setpoint tracking response in the frame of direct synthesis approach involving single tunable parameter. The tunable parameter’s values are obtained so that the robust stability condition is satisfied along with the justification of robust performance constraint in the presence of unstructured uncertainties. The effectiveness of the proposed method is verified on different plants and its advantages over some existing methods are also highlighted. Suggested approach yields improved performance with less control effort and enhanced robustness.

This paper presents a coordinated design of a model predictive controller (MPC) and switching law for delayed nonlinear switched systems with Lipschitz property. The article derives delay-dependent recursive feasibility constraints to handle disturbances in a polytopic form and incorporates them into the optimization problem. The control gain at each step is determined to ensure the feasibility of the optimization problem during the execution time of each sub-system. Additionally, constraints related to the cost function and H$\infty$ performance conditions are introduced in the developed linear matrix inequality problem to minimize the cost function with an infinite predictive horizon and guarantee controller robustness against unknown constant delays. The coordinated design of the MPC and switching law is achieved using the multiple Lyapunov–Krasovskii functional, which reduces the strictness of the controller constraints compared to the switched Lyapunov–Krasovskii functional. However, it imposes a dwell-time limitation on the switching law. By adopting the PDT structure, the dwell-time limitation is reduced compared to common structures. To evaluate the proposed design, it is applied to a water pollution system, and its performance is assessed. The results demonstrate superior performance compared to previous works.

The delta operator modeling provides a unified framework for both continuous-time and discrete-time modeling in system theory. At high sampling rate, the shift operator fails to provide meaningful information whereas, the delta operator parameterized system provides the same results as of continuous time systems. In this paper reduced order modeling of delta operator parameterized systems is considered. A complex domain ($\delta$) optimal frequency matching (OFM) technique is proposed and frequency points are optimized using Particle Swarm Optimization (PSO) algorithm. This OFM is then utilized to find the reduced order model of the higher order system. PSO algorithm is a robust, global optimization technique, used to find these OFMs and thereby used to find the coefficients of the reduced order model by minimizing a cost function developed based on the responses of the higher order model and that of the reduced order model when both are excited by pseudo random binary sequences (PRBS). The performance characteristics are evaluated in software simulation using MATLAB considering example of higher order system in delta domain and time & frequency responses of the corresponding reduced model.