IFAC Journal of Systems and Control最新文献

英文中文

Compatible realisation of control and identification of direct adaptive control via probing signal auto-elimination 通过探测信号自动消去实现直接自适应控制的兼容实现和辨识

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-30 DOI: 10.1016/j.ifacsc.2026.100375

Akira Takakura , Takashi Yokoyama , Takahiro Nozaki , Shuichi Adachi , Hiromitsu Ohmori

The model reference adaptive control system is an adaptive controller that maintains the control performance even when the uncertainty of the controlled system’s parameters is high, and its design methodology is well established. In particular, the direct MRACS excels in responsiveness; however, it suffers from the problem that its adjustable parameters do not converge to their true values. To converge the adjustable parameters to their true values, a conventional method involves injecting a probing signal to satisfy the PE property; however, this compromises the control performance. Thus, a control error-based probing signal auto-elimination scheme is proposed in this study, which adaptively regulates the probing signal based solely on the control error without predefined elimination timing. This enables the identification of adjustable parameters during transient phases, while automatically suppressing the probing signal once sufficient tracking performance is achieved. Furthermore, unlike existing probing-based methods, the proposed scheme allows re-injection of the probing signal when performance degradation is detected, thereby achieving a compatible realisation of identification and control within a single framework. Therefore, the proposed scheme simultaneously contributes to the identification and control, significantly reducing the tracking error. The validity of the proposed structure was confirmed by simulations under plant variation conditions.

模型参考自适应控制系统是在被控系统参数不确定性较大时仍能保持控制性能的一种自适应控制器，其设计方法是成熟的。特别是，直接的MRACS在反应性方面表现出色；然而，它的可调参数不收敛于其真实值的问题。为了使可调参数收敛到它们的真值，传统的方法是注入探测信号以满足PE特性；然而，这损害了控制性能。因此，本文提出了一种基于控制误差的探测信号自动消除方案，该方案仅根据控制误差对探测信号进行自适应调节，无需预先设定消除时间。这样可以在瞬态阶段识别可调参数，同时一旦达到足够的跟踪性能就自动抑制探测信号。此外，与现有的基于探测的方法不同，该方案允许在检测到性能下降时重新注入探测信号，从而在单个框架内实现识别和控制的兼容实现。因此，所提出的方案同时有助于识别和控制，大大减少了跟踪误差。在植物变异条件下的仿真验证了该结构的有效性。

{"title":"Compatible realisation of control and identification of direct adaptive control via probing signal auto-elimination","authors":"Akira Takakura , Takashi Yokoyama , Takahiro Nozaki , Shuichi Adachi , Hiromitsu Ohmori","doi":"10.1016/j.ifacsc.2026.100375","DOIUrl":"10.1016/j.ifacsc.2026.100375","url":null,"abstract":"<div><div>The model reference adaptive control system is an adaptive controller that maintains the control performance even when the uncertainty of the controlled system’s parameters is high, and its design methodology is well established. In particular, the direct MRACS excels in responsiveness; however, it suffers from the problem that its adjustable parameters do not converge to their true values. To converge the adjustable parameters to their true values, a conventional method involves injecting a probing signal to satisfy the PE property; however, this compromises the control performance. Thus, a control error-based probing signal auto-elimination scheme is proposed in this study, which adaptively regulates the probing signal based solely on the control error without predefined elimination timing. This enables the identification of adjustable parameters during transient phases, while automatically suppressing the probing signal once sufficient tracking performance is achieved. Furthermore, unlike existing probing-based methods, the proposed scheme allows re-injection of the probing signal when performance degradation is detected, thereby achieving a compatible realisation of identification and control within a single framework. Therefore, the proposed scheme simultaneously contributes to the identification and control, significantly reducing the tracking error. The validity of the proposed structure was confirmed by simulations under plant variation conditions.</div></div>","PeriodicalId":29926,"journal":{"name":"IFAC Journal of Systems and Control","volume":"35 ","pages":"Article 100375"},"PeriodicalIF":1.8,"publicationDate":"2026-01-30","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077720","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Multisine input signal design for constrained, “plant-friendly” system identification of nonlinear systems 多正弦输入信号的设计约束，“植物友好”系统辨识的非线性系统

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-29 DOI: 10.1016/j.ifacsc.2026.100371

Sarasij Banerjee , Eric Hekler , Daniel E. Rivera

This paper presents a methodology for optimizing “plant-friendly” multisine input signals to identify nonlinear dynamic systems under time-domain input and output constraints, without requiring a global parametric model a priori. The goal is to construct an informative dataset for open-loop, data-driven identification while selecting operational requirements. A weighted optimization framework is proposed to minimize the output crest factor resulting from a data-driven model, with penalties for violating input and output constraints. Model-on-Demand (MoD) estimation is employed to simulate outputs using prior data, effectively predicting nonlinear responses without global modeling. This MoD-based formulation enables evaluating output crest factors and output constraint compliance with modest modeling effort and improved impact. The resulting non-smooth, non-convex problem is solved using the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm, which perturbs the multisine phase vector to achieve the desired performance efficiently. This method supports the concept of identification test monitoring, as illustrated in this paper. Within the identification test loops, each optimized excitation is applied to gather new estimation data, iteratively refining MoD-based output predictions and improving constraint satisfaction. The method’s effectiveness is demonstrated through a safety-critical case study on a Susceptible-Infected-Recovered (SIR) epidemiological network, showing that the optimized excitation yields highly informative data for identification while keeping the infection spread within safe limits.

本文提出了一种优化“植物友好”多正弦输入信号的方法，以识别时域输入和输出约束下的非线性动态系统，而不需要先验的全局参数模型。目标是在选择操作需求的同时，为开环、数据驱动的识别构建信息数据集。提出了一种加权优化框架，以最小化由数据驱动模型产生的输出波峰因子，并对违反输入和输出约束进行惩罚。模型-按需（MoD）估计采用先验数据模拟输出，有效预测非线性响应而无需全局建模。这种基于模型的公式可以通过适度的建模努力和改进的影响来评估输出峰值因子和输出约束依从性。采用同步摄动随机逼近（SPSA）算法对多正弦相位矢量进行摄动以有效地达到预期的性能。该方法支持识别测试监控的概念，如本文所示。在识别测试循环中，应用每个优化的激励来收集新的估计数据，迭代地改进基于mod的输出预测，提高约束满意度。通过对易感-感染-恢复（SIR）流行病学网络的安全关键案例研究，证明了该方法的有效性，表明优化的激励产生了用于识别的高信息量数据，同时将感染传播保持在安全范围内。

{"title":"Multisine input signal design for constrained, “plant-friendly” system identification of nonlinear systems","authors":"Sarasij Banerjee , Eric Hekler , Daniel E. Rivera","doi":"10.1016/j.ifacsc.2026.100371","DOIUrl":"10.1016/j.ifacsc.2026.100371","url":null,"abstract":"<div><div>This paper presents a methodology for optimizing “plant-friendly” multisine input signals to identify nonlinear dynamic systems under time-domain input and output constraints, without requiring a global parametric model <em>a priori</em>. The goal is to construct an informative dataset for open-loop, data-driven identification while selecting operational requirements. A weighted optimization framework is proposed to minimize the output crest factor resulting from a data-driven model, with penalties for violating input and output constraints. Model-on-Demand (MoD) estimation is employed to simulate outputs using prior data, effectively predicting nonlinear responses without global modeling. This MoD-based formulation enables evaluating output crest factors and output constraint compliance with modest modeling effort and improved impact. The resulting non-smooth, non-convex problem is solved using the Simultaneous Perturbation Stochastic Approximation (SPSA) algorithm, which perturbs the multisine phase vector to achieve the desired performance efficiently. This method supports the concept of <em>identification test monitoring</em>, as illustrated in this paper. Within the identification test loops, each optimized excitation is applied to gather new estimation data, iteratively refining MoD-based output predictions and improving constraint satisfaction. The method’s effectiveness is demonstrated through a safety-critical case study on a Susceptible-Infected-Recovered (SIR) epidemiological network, showing that the optimized excitation yields highly informative data for identification while keeping the infection spread within safe limits.</div></div>","PeriodicalId":29926,"journal":{"name":"IFAC Journal of Systems and Control","volume":"35 ","pages":"Article 100371"},"PeriodicalIF":1.8,"publicationDate":"2026-01-29","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Data-driven design of dynamic quantizers applicable to nonminimum phase systems 非最小相位系统动态量化器的数据驱动设计

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-23 DOI: 10.1016/j.ifacsc.2026.100369

Yusuke Fujimoto , Yuki Minami

This paper discusses the data-driven design of a dynamic quantizer for control systems with discrete-valued input. We consider a quantizer with a noise-shaping filter that converts the continuous-valued input into the discrete-valued input, and discuss how to optimize the filter to minimize the error between the system outputs with and without quantization. It is known that this output deterioration can be measured by the

H_{\infty}

norm of a transfer function that depends on both the system and the noise-shaping filter. This paper focuses on data-driven estimation of the

H_{\infty}

norm from its input–output data, and virtually constructs input–output data for the transfer function. Then the output deterioration is minimized by minimizing this

H_{\infty}

norm. The effectiveness of the proposed approach is demonstrated through a numerical example.

本文讨论了具有离散值输入的控制系统动态量化器的数据驱动设计。我们考虑了一个带有噪声整形滤波器的量化器，它可以将连续值输入转换为离散值输入，并讨论了如何优化滤波器以最小化有量化和没有量化的系统输出之间的误差。众所周知，这种输出劣化可以通过传递函数的H∞范数来测量，该传递函数依赖于系统和噪声整形滤波器。本文从H∞范数的输入输出数据出发，研究了H∞范数的数据驱动估计，并虚拟构造了传递函数的输入输出数据。然后通过最小化这个H∞范数来最小化输出劣化。通过数值算例验证了该方法的有效性。

引用次数: 0

Regularized GLISp for sensor-guided human-in-the-loop optimization 用于传感器引导的人在环优化的正则化GLISp

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-22 DOI: 10.1016/j.ifacsc.2026.100368

Matteo Cercola , Michele Lomuscio , Dario Piga , Simone Formentin

Human-in-the-loop calibration is often addressed via preference-based optimization, where algorithms learn from pairwise comparisons rather than explicit cost evaluations. While effective, methods such as Preferential Bayesian Optimization or Global optimization based on active preference learning with radial basis functions (GLISp) treat the system as a black box and ignore informative sensor measurements. In this work, we introduce a sensor-guided regularized extension of GLISp that integrates measurable descriptors into the preference-learning loop through a physics-informed hypothesis function and a least-squares regularization term. This injects grey-box structure, combining subjective feedback with quantitative sensor information while preserving the flexibility of preference-based search. Numerical evaluations on an analytical benchmark and on a human-in-the-loop vehicle suspension tuning task show faster convergence and superior final solutions compared to baseline GLISp.

人在环校准通常通过基于偏好的优化来解决，算法从两两比较中学习，而不是明确的成本评估。虽然有效，但诸如优先贝叶斯优化或基于径向基函数主动偏好学习（GLISp）的全局优化等方法将系统视为黑箱，忽略了信息传感器测量。在这项工作中，我们引入了GLISp的传感器引导正则化扩展，该扩展通过物理信息假设函数和最小二乘正则化项将可测量描述符集成到偏好学习循环中。这注入了灰盒结构，将主观反馈与定量传感器信息相结合，同时保持了基于偏好的搜索的灵活性。对分析基准和人在环车辆悬架调整任务的数值评估表明，与基线GLISp相比，该方法收敛速度更快，最终解决方案更优。

引用次数: 0

Stability-constrained policy optimization under unknown rewards 未知奖励下的稳定约束策略优化

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-21 DOI: 10.1016/j.ifacsc.2026.100366

Thomas Banker, Nathan P. Lawrence, Ali Mesbah

A major challenge in reinforcement learning (RL) is guaranteeing an agent’s closed-loop stability under unknown, possibly sparse, reward functions. While model-free RL is flexible to a variety of systems and rewards, model-based control strategies such as optimization-based control naturally accommodate prior system models to provide guarantees on safety and stability. However, these models may not be representative of the true global performance objective, resulting in suboptimal policies. In this paper, we present a policy search RL approach that decouples the stability requirement from the global performance objective. The key idea is to use an optimization-based policy structure as an effective stabilizing parameterization with which the agent can learn to maximize an unknown reward in a model-free fashion. Specifically, the agent employs a predictive control architecture and implicitly learns a stabilizing terminal cost, which is constructed through fixed-point iterations of the discrete algebraic Riccati equation. By implicitly differentiating this fixed-point, derivatives of the stability condition inform policy gradients. The proposed approach is shown to design high-performance, stabilizing policies for various sparse, global performance objectives. Furthermore, the proposed approach can account for uncertainty in the dynamics using the stochastic discrete algebraic Riccati equation to promote robust stability. This work demonstrates a principled policy search RL approach, integrating prior models and system observations in an agent’s design, towards safe and reliable decision-making under uncertainty.

强化学习（RL）的一个主要挑战是保证智能体在未知的、可能稀疏的奖励函数下的闭环稳定性。虽然无模型强化学习对各种系统和奖励都很灵活，但基于模型的控制策略（如基于优化的控制）自然地适应了先前的系统模型，以提供安全性和稳定性的保证。然而，这些模型可能不能代表真正的全局性能目标，从而导致次优策略。在本文中，我们提出了一种策略搜索RL方法，该方法将稳定性要求与全局性能目标解耦。关键思想是使用基于优化的策略结构作为有效的稳定参数化，通过该参数化，智能体可以学习以无模型的方式最大化未知奖励。具体而言，该智能体采用预测控制体系结构，通过对离散代数Riccati方程的不动点迭代构造一个稳定的终端代价，并隐式学习。通过隐式微分这个不动点，稳定性条件的导数告知政策梯度。所提出的方法被证明可以为各种稀疏的全局性能目标设计高性能、稳定的策略。此外，该方法可以利用随机离散代数Riccati方程来解释动力学中的不确定性，从而提高鲁棒稳定性。这项工作展示了一种原则性的策略搜索强化学习方法，在智能体设计中集成了先前的模型和系统观察，以实现不确定性下的安全可靠决策。

{"title":"Stability-constrained policy optimization under unknown rewards","authors":"Thomas Banker, Nathan P. Lawrence, Ali Mesbah","doi":"10.1016/j.ifacsc.2026.100366","DOIUrl":"10.1016/j.ifacsc.2026.100366","url":null,"abstract":"<div><div>A major challenge in reinforcement learning (RL) is guaranteeing an agent’s closed-loop stability under unknown, possibly sparse, reward functions. While model-free RL is flexible to a variety of systems and rewards, model-based control strategies such as optimization-based control naturally accommodate prior system models to provide guarantees on safety and stability. However, these models may not be representative of the true global performance objective, resulting in suboptimal policies. In this paper, we present a policy search RL approach that decouples the stability requirement from the global performance objective. The key idea is to use an optimization-based policy structure as an effective stabilizing parameterization with which the agent can learn to maximize an unknown reward in a model-free fashion. Specifically, the agent employs a predictive control architecture and implicitly learns a stabilizing terminal cost, which is constructed through fixed-point iterations of the discrete algebraic Riccati equation. By implicitly differentiating this fixed-point, derivatives of the stability condition inform policy gradients. The proposed approach is shown to design high-performance, stabilizing policies for various sparse, global performance objectives. Furthermore, the proposed approach can account for uncertainty in the dynamics using the stochastic discrete algebraic Riccati equation to promote robust stability. This work demonstrates a principled policy search RL approach, integrating prior models and system observations in an agent’s design, towards safe and reliable decision-making under uncertainty.</div></div>","PeriodicalId":29926,"journal":{"name":"IFAC Journal of Systems and Control","volume":"35 ","pages":"Article 100366"},"PeriodicalIF":1.8,"publicationDate":"2026-01-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"146077767","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

On continuous-time sparse identification of nonlinear polynomial systems 非线性多项式系统的连续时间稀疏辨识

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-15 DOI: 10.1016/j.ifacsc.2026.100365

Mazen Alamir

This paper leverages recent advances in high derivatives reconstruction from noisy-time series and sparse multivariate polynomial identification in order to improve the process of parsimoniously identifying, from a small amount of data, unknown Single-Input/Single-Output nonlinear dynamics of relative degree up to 4. The methodology is illustrated on the Electronic Throttle Controlled automotive system.

本文利用噪声时间序列的高导数重构和稀疏多元多项式辨识的最新进展，改进了从少量数据中简化识别相对程度高达4的未知单输入/单输出非线性动力学的过程。以电子节气门控制汽车系统为例说明了该方法。

引用次数: 0

Joint state and parameter estimation in quantum systems using cubature Kalman filtering 利用培养卡尔曼滤波的量子系统联合状态和参数估计

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-14 DOI: 10.1016/j.ifacsc.2026.100363

Eram Taslima, Shyam Kamal, R.K. Saket

This paper addresses the challenge of state estimation for two-level quantum systems governed by stochastic master equations, particularly when key Hamiltonian parameters are unknown. The critical parameters such as the qubit resonance frequency and the decay rate play a crucial role in determining system dynamics, hence their accurate estimation is essential for reliable state reconstruction. A robust framework based on the cubature Kalman filter (CKF) is developed that effectively handles both correlated and decorrelated noise processes inherent to quantum homodyne measurement. The proposed approach effectively mitigates performance degradation caused by parametric uncertainty, providing enhanced adaptability and robustness. Numerical simulations on a qubit in a cavity show that the CKF-based method achieves better estimation accuracy and faster convergence compared to the extended Kalman filter.

本文讨论了由随机主方程控制的二能级量子系统的状态估计问题，特别是在关键哈密顿参数未知的情况下。量子比特共振频率和衰减率等关键参数在系统动力学中起着至关重要的作用，因此它们的准确估计对于可靠的状态重建至关重要。提出了一种基于稳态卡尔曼滤波（CKF）的鲁棒框架，该框架能有效地处理量子同差测量中固有的相关和去相关噪声过程。该方法有效地缓解了参数不确定性引起的性能下降，增强了自适应性和鲁棒性。在腔内量子比特上的数值模拟表明，与扩展卡尔曼滤波相比，基于ckf的方法具有更好的估计精度和更快的收敛速度。

引用次数: 0

Symbolic learning of interpretable reduced-order models for jumping quadruped robots 跳跃四足机器人可解释降阶模型的符号学习

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-08 DOI: 10.1016/j.ifacsc.2025.100360

Gioele Buriani , Jingyue Liu , Maximilian Stölzle , Cosimo Della Santina , Jiatao Ding

Reduced-order models are central to motion planning and control of quadruped robots, yet existing templates are often hand-crafted for a specific locomotion modality. This motivates the need for automatic methods that extract task-specific, interpretable low-dimensional dynamics directly from data. We propose a methodology that combines a linear autoencoder with symbolic regression to derive such models. The linear autoencoder provides a consistent latent embedding for configurations, velocities, accelerations, and inputs, enabling the sparse identification of nonlinear dynamics (SINDy) to operate in a compact, physics-aligned space. A multi-phase, hybrid-aware training scheme ensures coherent latent coordinates across contact transitions. We focus our validation on quadruped jumping—a representative, challenging, yet contained scenario in which a principled template model is especially valuable. The resulting symbolic dynamics outperform the state-of-the-art handcrafted actuated spring-loaded inverted pendulum (aSLIP) baseline in simulation and hardware across multiple robots and jumping modalities.

降阶模型是四足机器人运动规划和控制的核心，但现有的模板通常是针对特定的运动模式手工制作的。这激发了对直接从数据中提取任务特定的、可解释的低维动态的自动方法的需求。我们提出了一种结合线性自编码器和符号回归的方法来推导这样的模型。线性自编码器为配置、速度、加速度和输入提供一致的潜在嵌入，使非线性动力学（SINDy）的稀疏识别能够在紧凑的物理对齐空间中运行。一个多阶段，混合感知的训练方案，确保在接触过渡连贯的潜在坐标。我们将验证重点放在四足跳跃上——这是一个具有代表性的、具有挑战性的、但包含的场景，在这个场景中，有原则的模板模型特别有价值。由此产生的符号动力学在多个机器人和跳跃模式的仿真和硬件方面优于最先进的手工制作驱动弹簧加载倒立摆（aSLIP）基线。

{"title":"Symbolic learning of interpretable reduced-order models for jumping quadruped robots","authors":"Gioele Buriani , Jingyue Liu , Maximilian Stölzle , Cosimo Della Santina , Jiatao Ding","doi":"10.1016/j.ifacsc.2025.100360","DOIUrl":"10.1016/j.ifacsc.2025.100360","url":null,"abstract":"<div><div>Reduced-order models are central to motion planning and control of quadruped robots, yet existing templates are often hand-crafted for a specific locomotion modality. This motivates the need for automatic methods that extract task-specific, interpretable low-dimensional dynamics directly from data. We propose a methodology that combines a linear autoencoder with symbolic regression to derive such models. The linear autoencoder provides a consistent latent embedding for configurations, velocities, accelerations, and inputs, enabling the sparse identification of nonlinear dynamics (SINDy) to operate in a compact, physics-aligned space. A multi-phase, hybrid-aware training scheme ensures coherent latent coordinates across contact transitions. We focus our validation on quadruped jumping—a representative, challenging, yet contained scenario in which a principled template model is especially valuable. The resulting symbolic dynamics outperform the state-of-the-art handcrafted actuated spring-loaded inverted pendulum (aSLIP) baseline in simulation and hardware across multiple robots and jumping modalities.</div></div>","PeriodicalId":29926,"journal":{"name":"IFAC Journal of Systems and Control","volume":"35 ","pages":"Article 100360"},"PeriodicalIF":1.8,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145925981","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Ensemble self-training deep partial least squares models for stable semi-supervised predictive learning and data analytics 用于稳定半监督预测学习和数据分析的集成自训练深度偏最小二乘模型

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-08 DOI: 10.1016/j.ifacsc.2026.100362

Junhua Zheng , Zhiqiang Ge , Li Sun

While deep learning has made significant achievements in the past years, it suffers from several serious shortcomings. Particularly, the performance of deep learning may be severely degraded under a small size of labeled training dataset, the case of which is quite common in industrial application scenarios although we are in the age of big data. In this paper, a semi-supervised deep model is proposed for predictive learning and data analytics, which is based upon the recently developed lightweight deep partial least squares model (PLS) structure. Precisely, the simple self-training strategy is used as the driving force to formulate the semi-supervised deep PLS model, which has no restriction in model structure and thus is flexible for predictive learning. In addition, to reduce the uncertainty of the self-training process, i.e. prediction error accumulation, different random seeds are introduced for model training, the results of which are combined together through an ensemble learning strategy. As a result, the predictive model becomes more stable and robust to those uncertainties introduced by both unlabeled data and the semi-supervised learning process. A real industrial example is provided for performance evaluation of the proposed method.

虽然深度学习在过去几年取得了重大成就，但它也存在一些严重的缺点。特别是在小规模的标记训练数据集下，深度学习的性能可能会严重下降，虽然我们处于大数据时代，但这种情况在工业应用场景中很常见。本文基于近年来发展起来的轻量级深度偏最小二乘模型（PLS）结构，提出了一种用于预测学习和数据分析的半监督深度模型。准确地说，利用简单的自我训练策略作为动力来构建半监督深度PLS模型，该模型不受模型结构的限制，具有预测学习的灵活性。此外，为了减少自训练过程的不确定性，即预测误差积累，引入了不同的随机种子进行模型训练，并通过集成学习策略将其结果组合在一起。因此，该预测模型对于未标记数据和半监督学习过程引入的不确定性变得更加稳定和鲁棒。给出了一个实际的工业实例，对该方法进行了性能评价。

{"title":"Ensemble self-training deep partial least squares models for stable semi-supervised predictive learning and data analytics","authors":"Junhua Zheng , Zhiqiang Ge , Li Sun","doi":"10.1016/j.ifacsc.2026.100362","DOIUrl":"10.1016/j.ifacsc.2026.100362","url":null,"abstract":"<div><div>While deep learning has made significant achievements in the past years, it suffers from several serious shortcomings. Particularly, the performance of deep learning may be severely degraded under a small size of labeled training dataset, the case of which is quite common in industrial application scenarios although we are in the age of big data. In this paper, a semi-supervised deep model is proposed for predictive learning and data analytics, which is based upon the recently developed lightweight deep partial least squares model (PLS) structure. Precisely, the simple self-training strategy is used as the driving force to formulate the semi-supervised deep PLS model, which has no restriction in model structure and thus is flexible for predictive learning. In addition, to reduce the uncertainty of the self-training process, i.e. prediction error accumulation, different random seeds are introduced for model training, the results of which are combined together through an ensemble learning strategy. As a result, the predictive model becomes more stable and robust to those uncertainties introduced by both unlabeled data and the semi-supervised learning process. A real industrial example is provided for performance evaluation of the proposed method.</div></div>","PeriodicalId":29926,"journal":{"name":"IFAC Journal of Systems and Control","volume":"35 ","pages":"Article 100362"},"PeriodicalIF":1.8,"publicationDate":"2026-01-08","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145977242","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

Convex computation of regions of attraction from data using sums-of-squares programming 利用平方和规划从数据中凸计算吸引区域

IF 1.8 Q3 AUTOMATION & CONTROL SYSTEMS

IFAC Journal of Systems and Control

Pub Date : 2026-01-06 DOI: 10.1016/j.ifacsc.2026.100361

Oumayma Khattabi , Matteo Tacchi-Bénard , Sorin Olaru

This paper focuses on the analysis of the Region of Attraction (RoA) for unknown autonomous dynamical systems. A data-driven approach based on the moment-Sum of Squares (SoS) hierarchy is proposed, enabling novel RoA outer approximations despite the reduced information on the dynamics. The main contribution consists of bypassing the system model and, hence, the recurring constraint on its polynomial structure. Numerical experiments showcase the influence of data on learned approximating sets, highlighting the potential of this method.

本文主要研究未知自主动力系统的吸引区（RoA）问题。提出了一种基于矩平方和（so）层次结构的数据驱动方法，尽管减少了动力学信息，但仍能实现新的RoA外部逼近。其主要贡献在于绕过了系统模型，从而避免了对其多项式结构的反复约束。数值实验显示了数据对学习近似集的影响，突出了该方法的潜力。

引用次数: 0

下一页尾页

类型

全部化学•材料生命科学医学物理工程技术环境•农林材料科学地球科学法学管理学化学环境科学与生态学计算机科学教育学经济学农林科学人文科学生物学数学物理与天体物理心理学综合性期刊其他工业工程理学历史学农学文学信息工程

数据库

全部 ACS Publications Elsevier ieeexplore Springer The Royal Society of Chemistry Wiley

期刊

IFAC Journal of Systems and Control

全部 Acc. Chem. Res. ACS Applied Bio Materials ACS Appl. Electron. Mater. ACS Appl. Energy Mater. ACS Appl. Mater. Interfaces ACS Appl. Nano Mater. ACS Appl. Polym. Mater. ACS BIOMATER-SCI ENG ACS Catal. ACS Cent. Sci. ACS Chem. Biol. ACS Chemical Health & Safety ACS Chem. Neurosci. ACS Comb. Sci. ACS Earth Space Chem. ACS Energy Lett. ACS Infect. Dis. ACS Macro Lett. ACS Mater. Lett. ACS Med. Chem. Lett. ACS Nano ACS Omega ACS Photonics ACS Sens. ACS Sustainable Chem. Eng. ACS Synth. Biol. Anal. Chem. BIOCHEMISTRY-US Bioconjugate Chem. BIOMACROMOLECULES Chem. Res. Toxicol. Chem. Rev. Chem. Mater. CRYST GROWTH DES ENERG FUEL Environ. Sci. Technol. Environ. Sci. Technol. Lett. Eur. J. Inorg. Chem. IND ENG CHEM RES Inorg. Chem. J. Agric. Food. Chem. J. Chem. Eng. Data J. Chem. Educ. J. Chem. Inf. Model. J. Chem. Theory Comput. J. Med. Chem. J. Nat. Prod. J PROTEOME RES J. Am. Chem. Soc. LANGMUIR MACROMOLECULES Mol. Pharmaceutics Nano Lett. Org. Lett. ORG PROCESS RES DEV ORGANOMETALLICS J. Org. Chem. J. Phys. Chem. J. Phys. Chem. A J. Phys. Chem. B J. Phys. Chem. C J. Phys. Chem. Lett. Analyst Anal. Methods Biomater. Sci. Catal. Sci. Technol. Chem. Commun. Chem. Soc. Rev. CHEM EDUC RES PRACT CRYSTENGCOMM Dalton Trans. Energy Environ. Sci. ENVIRON SCI-NANO ENVIRON SCI-PROC IMP ENVIRON SCI-WAT RES Faraday Discuss. Food Funct. Green Chem. Inorg. Chem. Front. Integr. Biol. J. Anal. At. Spectrom. J. Mater. Chem. A J. Mater. Chem. B J. Mater. Chem. C Lab Chip Mater. Chem. Front. Mater. Horiz. MEDCHEMCOMM Metallomics Mol. Biosyst. Mol. Syst. Des. Eng. Nanoscale Nanoscale Horiz. Nat. Prod. Rep. New J. Chem. Org. Biomol. Chem. Org. Chem. Front. PHOTOCH PHOTOBIO SCI PCCP Polym. Chem.

﹀