Article
Automation & Control Systems
Huafeng Xia
Summary: This article presents a new iterative method for parameter estimation of a class of closed-loop Hammerstein systems. The method solves the crossproducts between the parameters of the linear block and the nonlinear block using the key term separation technique. By utilizing the hierarchical identification principle, the system is decomposed into two subidentification models, reducing computational complexity. The method also maximizes the maximum likelihood cost function by using input and output data with a data window, leading to improved parameter estimation accuracy. Numerical simulation and experimental results demonstrate the effectiveness of the proposed algorithm.
INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL
(2023)
Article
Automation & Control Systems
Cun Wang, Zupeng Zhou, Xufeng Liu
Summary: This paper investigates the consensus control problem of multi-agent systems and proposes a distributed closed-loop accelerated iterative learning control protocol based on the network topology and the output form of each follower to speed up the convergence rate of the consensus error. Numerical simulations demonstrate the effectiveness of the proposed protocol.
INTERNATIONAL JOURNAL OF ROBUST AND NONLINEAR CONTROL
(2023)
Review
Endocrinology & Metabolism
Charlotte K. Boughton, Roman Hovorka
Summary: Advances in diabetes technologies have led to the development of automated closed-loop insulin delivery systems, which are changing the way type 1 diabetes is managed. This review examines the evidence supporting currently available closed-loop systems and those in development, including DIY systems, as well as the challenges and potential enhancements for clinical adoption.
Article
Automation & Control Systems
Kuan Li, Hao Luo, Yuchen Jiang, Dejia Tang, Hongyan Yang
Summary: This paper presents a subspace identification method for closed-loop systems with unknown deterministic disturbances. By constructing the row space to approximate the unknown disturbances, effective identification of the closed-loop systems can be achieved.
IEEE-CAA JOURNAL OF AUTOMATICA SINICA
(2023)
Article
Automation & Control Systems
Kuan Li, Hao Luo, Shen Yin, Okyay Kaynak
Summary: This article introduces a novel data-driven bias-eliminated subspace identification approach for closed-loop systems, which effectively eliminates the pole estimation bias caused by the correlation between inputs and noise under feedback control. The proposed method establishes a general framework applicable to both open-loop and closed-loop system identification, and outperforms other closed-loop methods in terms of performance comparisons. The effectiveness of the identified system is further demonstrated in the vehicle lateral dynamic system.
IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS
(2021)
Article
Green & Sustainable Science & Technology
Wanju Yuan, Zhuoheng Chen, Stephen E. Grasby, Edward Little
Summary: Closed-loop geothermal systems can provide reasonable temperature and heat duty for over 30 years with multiple laterals, but their sensitivity to reservoir thermal conductivity controls outlet temperature and interference between wells over time, while resident time in the horizontal section dictates heat transport efficiency.
Article
Engineering, Electrical & Electronic
Holger Boche, Volker Pohl
Summary: This paper proves that simulating non-trivial, linear time-invariant (LTI) systems of the first order on a digital computer leads to complexity blowup. It also demonstrates that the same complexity blowup occurs when simulating higher-order LTI systems with real poles. Additionally, calculations of Fourier series approximations and Fourier transforms also experience similar complexity blowup.
IEEE TRANSACTIONS ON SIGNAL PROCESSING
(2021)
Article
Computer Science, Information Systems
Abdelali El Aroudi, Luis Benadero, Reham Haroun, Luis Martinez-Salamero, Chi K. Tse
Summary: Constant power loading (CPL) can lead to instabilities in open-loop power converters operating in discontinuous conduction mode (DCM). This study shows that the behavior of an open-loop boost converter under DCM operation, loaded with a tightly regulated buck converter, is similar to when it is loaded by an ideal CPL. Bifurcation phenomena were observed in three elementary open-loop DC-DC converters operating in DCM. It was found that the period-doubling route to chaos in the DC-DC boost converter is interrupted by dangerous destructive dynamics due to the unlimited load current in the CPL.
Article
Engineering, Mechanical
Yongjun Pan, Wei Dai, Liming Huang, Zhixiong Li, Aki Mikkola
Summary: This paper presents an iterative refinement algorithm to accelerate the numerical integration process of a semi-recursive multibody formulation. By reusing the constraint Jacobian matrix factorization and generalized mass matrix factorization, the algorithm enhances the solution of dependent relative velocities and independent relative accelerations.
MECHANICAL SYSTEMS AND SIGNAL PROCESSING
(2021)
Article
Engineering, Electrical & Electronic
Giacomo Innocenti, Alberto Tesi, Mauro Di Marco, Mauro Forti
Summary: This paper investigates the possibility of embedding known attractors into memristor circuits, replicating the dynamics of nonlinear systems with complex attractors in input-less circuits by selecting suitable parameters. The results show that the dynamics of the nonlinear system can be replicated in a memristor circuit under constant input, and there is a one-one correspondence between the dynamics of the system and that displayed on the invariant manifolds of the memristor circuit.
IEEE JOURNAL ON EMERGING AND SELECTED TOPICS IN CIRCUITS AND SYSTEMS
(2022)
Review
Endocrinology & Metabolism
Sophie Templer
Summary: Closed-loop systems for automated insulin delivery have significantly advanced in recent years, from bulky, bedside technologies to miniaturized, wearable devices. These systems use interstitial glucose sensing, subcutaneous insulin pumps, and sophisticated algorithms. Future challenges include the development of fully closed-loop systems and the addition of glucagon to mitigate hypoglycemia risk.
FRONTIERS IN ENDOCRINOLOGY
(2022)
Article
Biochemical Research Methods
Suleman A. Naz, Van Thanh Huynh, Egan H. Doeven, Scott Adams, Abbas Kouzani, Rosanne M. Guijt
Summary: The adoption of mechatronic approaches has been driven by the demand for accurate control of the flowrate/pressure in chemical analytical systems. In the development of portable analytical devices, considering the instrument as a mechatronic system can help mitigate compromises made to decrease space, weight, or power consumption. Closed loop control systems have been effectively used to decrease the difference between desired and realized fluidic output. This review discusses the implementation of control systems for enhanced fluidic control, categorized by pump type, along with advanced control strategies and their implementation in portable analytical systems.
JOURNAL OF CHROMATOGRAPHY A
(2023)
Article
Engineering, Electrical & Electronic
Rakesh Ashok, Sana Naaz, Rashmi Kamran, Shalabh Gupta
Summary: Analog domain signal processing is an effective method for reducing power consumption in high capacity short reach coherent links. The demonstration of a carrier phase synchronization chip designed in a 130 nm SiGe BiCMOS technology has shown success in QPSK and 16-QAM modulation at 40 Gbps.
JOURNAL OF LIGHTWAVE TECHNOLOGY
(2021)
Article
Engineering, Multidisciplinary
Hong Jianwang, Ricardo A. Ramirez-Mendoza, Ruben Morales-Menendez
Summary: This paper introduces a method of combining iterative ideas with data-driven control to design both feedforward and feedback controllers simultaneously. By parameterizing two controllers and applying iterative ideas to identify parameter vectors, closed relations between controllers and expected transfer functions are derived. Equations for output-input sensitivity functions are also derived to aid in stability analysis.
MATHEMATICAL PROBLEMS IN ENGINEERING
(2021)
Article
Endocrinology & Metabolism
Giuseppe Papa, Rossella Cannarella, Rosita A. Condorelli, Concetta Finocchiaro, Aldo E. Calogero, Sandro La Vignera
Summary: This study suggests that switching to Hybrid Closed Loop systems can significantly improve glycometabolic outcomes for AID-na & iuml;ve T1D patients, allowing better control of blood glucose levels within the target range, and with high levels of satisfaction.
DIABETES RESEARCH AND CLINICAL PRACTICE
(2023)
Article
Engineering, Electrical & Electronic
Vikram Jain, Sebastian Giraldo, Jaro De Roose, Linyan Mei, Bert Boons, Marian Verhelst
Summary: This article presents TinyVers, a tiny versatile ultra-low power machine learning system-on-chip (SoC) designed for enhanced intelligence at the extreme edge. By utilizing dataflow reconfiguration and on-chip power management, TinyVers enables multi-modal support and efficient smart sensing applications. The SoC includes a RISC-V host processor, a dataflow reconfigurable ML accelerator, a deep sleep wake-up controller, and embedded magnetoresistive random access memory (eMRAM). It achieves high performance of up to 17.6 GOPS with a power consumption range from 1.7μW to 20mW, demonstrating flexibility and efficiency in various machine learning workloads.
IEEE JOURNAL OF SOLID-STATE CIRCUITS
(2023)
Article
Computer Science, Hardware & Architecture
Jun Yin, Marian Verhelst
Summary: This study proposes a robust sound source localization model for environments with noise and reverberation, which can run in real-time on edge devices with higher efficiency and smaller model size.
ACM TRANSACTIONS ON EMBEDDED COMPUTING SYSTEMS
(2023)
Correction
Biotechnology & Applied Microbiology
L. Van Peteghem, M. Sakarika, S. Matassa, K. Rabaey
APPLIED AND ENVIRONMENTAL MICROBIOLOGY
(2023)
Article
Engineering, Environmental
Eleni Stylianou, Jose M. Carvajal-Arroyo, Dimitrios Ladakis, Carol Sze Ki Lin, Vera Essmann, Sebastin Dorr, Jakob Marbach, Korneel Rabaey, Apostolis Koutinas, Chrysanthi Pateraki
Summary: A novel electrochemical membrane bioreactor (EMB) integrating succinic acid (SA) production and in situ separation was used with hydrolysates from the organic fraction of municipal solid waste (OFMSW) as feedstock. The fermentation process was optimized to achieve high SA productivity and efficiency, and the SA crystals were successfully purified for further applications. Scale-up of the process was influenced by yeast cell viability, cell design, and EMB configuration. Overall, this study demonstrated the potential of the EMB system for sustainable SA production and purification.
CHEMICAL ENGINEERING JOURNAL
(2023)
Article
Engineering, Environmental
Kai Cui, Kun Guo, Jose M. Carvajal-Arroyo, Jan Arends, Korneel Rabaey
Summary: Currently, most microbial electrosynthesis (MES) reactors are driven by biofilms, whether the cathode electron transfer is direct or indirect. The formation of cathode biofilms is time-consuming, and the current density achieved on these biofilms is relatively low. However, non-biofilm-driven MES faces the challenge of maintaining good Columbic efficiency (CE) at high current densities. In this study, a novel electro-H2 bubble column reactor with an external hollow fiber membrane gas-liquid contactor was developed to address this challenge. The reactor demonstrated efficient conversion of electrolytically-produced H2 and externally-supplied CO2 into acetate, with high acetate production rates and CE.
CHEMICAL ENGINEERING JOURNAL
(2023)
Article
Engineering, Electrical & Electronic
Koen Goetschalckx, Fengfeng Wu, Marian Verhelst
Summary: This article introduces a processor chip called DepFiN, which reduces energy costs by running CNNs in a depth-first execution mode. It also tackles the computational cost by supporting DW + PW layer pairs with a novel accelerator core. Benchmarking measurements show that the DepFiN chip achieves an efficiency of up to 8.2 TOPS/W on the MC-CNN-fast stereo-matching network and up to an 18-fold improvement is realized by supporting depth-first execution.
IEEE JOURNAL OF SOLID-STATE CIRCUITS
(2023)
Article
Engineering, Biomedical
Thomas Bos, Marian Verhelst, Wim Dehaene
Summary: This paper proposes a custom OFDM modem for ultrasound in-body communication, which can achieve efficient wireless communication under different channel conditions. Experimental results show that it can achieve a transmission rate of 470 kbps on a 14 cm thick piece of beef, with low power consumption.
IEEE TRANSACTIONS ON BIOMEDICAL CIRCUITS AND SYSTEMS
(2023)
Article
Computer Science, Hardware & Architecture
Steven Colleman, Man Shi, Marian Verhelst
Summary: This article presents COAC, a cross-layer framework that optimizes the flexibility of neural processing architectures by balancing configurability overhead against resulting energy and latency savings for end-to-end inference. COAC provides a systematical analysis of the architectural overhead and automates the search for the best unrolling combination(s) for efficient inference with limited hardware overhead.
IEEE TRANSACTIONS ON VERY LARGE SCALE INTEGRATION (VLSI) SYSTEMS
(2023)
Proceedings Paper
Computer Science, Hardware & Architecture
Arne Symons, Linyan Mei, Steven Colleman, Pouya Houshmand, Sebastian Karl, Marian Verhelst
Summary: To meet the increasing performance demand of DNN processing, specialized hardware accelerators are shifting towards multi-core architectures. Stream, an open-source design space exploration framework, co-optimizes HW architecture and fine-grained scheduling for multi-core DNN accelerators. It supports fine-grained layer fusion to optimize energy, latency, and on-chip memory footprint for constrained edge devices. Validation against three state-of-the-art chips and a case study on seven HW architectures demonstrate the reliability and capabilities of Stream. Results show significant improvements in HW efficiency under fine-grained scheduling, reducing the energy-delay product compared to traditional scheduling. Stream is open-source at github.com/ZigZag-Project/stream.
2023 IEEE INTERNATIONAL SYMPOSIUM ON PERFORMANCE ANALYSIS OF SYSTEMS AND SOFTWARE, ISPASS
(2023)
Proceedings Paper
Computer Science, Hardware & Architecture
Linyan Mei, Koen Goetschalckx, Arne Symons, Marian Verhelst
Summary: DNN workloads can be scheduled onto DNN accelerators in various ways, but existing cost models lack support for exploring the complete depth-first scheduling space. This work proposes a unified modeling framework called DeFiNES to fill in the gaps and analytically estimate the hardware cost for different schedules. The proposed model considers data access at every memory level and is validated against measured data, showing better solutions compared to existing state-of-the-art.
2023 IEEE INTERNATIONAL SYMPOSIUM ON HIGH-PERFORMANCE COMPUTER ARCHITECTURE, HPCA
(2023)
Article
Green & Sustainable Science & Technology
Xiaoting Zhang, Tyler Arbour, Daijun Zhang, Shiqiang Wei, Korneel Rabaey
Summary: Microbial electrosynthesis (MES) can produce multicarbon compounds from CO₂ using electricity as the energy source. This study assessed MES under 35-60 g/L salinity and found that acetate production decreased over time under high salinity conditions. Carbonate precipitation on cathodes inhibited CO₂ supply, electrode contact, and H₂ production, leading to reduced acetate production. By adjusting the levels of Ca²⁺ and Mg²⁺ in the reactors, stable acetate production was achieved. The research also highlighted the dominance of Acetobacterium and Sphaerochaeta bacteria in the MES system.
ENVIRONMENTAL SCIENCE AND ECOTECHNOLOGY
(2023)
Proceedings Paper
Engineering, Electrical & Electronic
Weijie Jiang, Pouya Houshmand, Marian Verhelst, Wim Dehaene
Summary: Digital in-memory compute (DIMC) is attracting attention for its robustness and ease of integration into practical tasks. However, existing DIMC macros have low area efficiency. This paper proposes a DIMC macro that uses unmodified 6T bit-cells with added MAC logic on the bit-line, and utilizes techniques to reduce energy penalty caused by increased bit-line loading.
IEEE 49TH EUROPEAN SOLID STATE CIRCUITS CONFERENCE, ESSCIRC 2023
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Josse Van Delm, Maarten Vandersteegenl, Alessio Burrello, Giuseppe Maria Sarda, Francesco Conti, Daniele Jahier Pagliari, Luca Benini, Marian Verhelst
Summary: The optimal deployment of deep neural networks on state-of-the-art Systems-on-Chips is crucial for tiny machine learning at the edge. We propose a compiler called HTVM that merges TVM with DORY to maximize the utilization of heterogeneous accelerators and minimize data movements.
2023 60TH ACM/IEEE DESIGN AUTOMATION CONFERENCE, DAC
(2023)
Proceedings Paper
Computer Science, Artificial Intelligence
Vikram Jain, Matheus Cavalcante, Nazareno Bruschi, Michael Rogenmoser, Thomas Benz, Andreas Kurth, Davide Rossi, Luca Benini, Marian Verhelst
Summary: This paper proposes an open-source AXI-compliant NoC fabric called PATRONoC to address the specific needs of multi-core DNN computing platforms. Evaluation results show that PATRONoC achieves higher area efficiency and throughput compared to classical NoCs.
2023 60TH ACM/IEEE DESIGN AUTOMATION CONFERENCE, DAC
(2023)
Proceedings Paper
Computer Science, Information Systems
Matteo Risso, Alessio Burrello, Giuseppe Maria Sarda, Luca Benini, Enrico Macii, Massimo Poncino, Marian Verhelst, Daniele Jahier Pagliari
Summary: In this study, we propose ODiMO, a hardware-aware tool for fine-grain mapping of DNNs across multiple accelerators onchip to reduce inference energy consumption or latency, while maintaining accuracy. Experimental results demonstrate that ODiMO effectively reduces energy/latency with limited accuracy drop compared to manual heuristic mappings.
2023 IEEE/ACM INTERNATIONAL SYMPOSIUM ON LOW POWER ELECTRONICS AND DESIGN, ISLPED
(2023)