Special Session 196: Control-Theoretic Analysis of Nonlinear and Uncertain Systems with Data-Driven and Learning Mechanisms

Adaptive Event-Triggered Control for High-Order Nonlinear Impulsive Systems Subjects to Parametric Uncertainty and External Disturbance

Yuanen Li Sun Yat-Sen University Peoples Rep of China Co-Author(s):    Yuanen Li, Xuefang Li, Wanquan Liu

Abstract: This paper investigates adaptive event-triggered control for high-order nonlinear impulsive systems with parametric uncertainty and external disturbance. Existing stabilization methods for nonlinear impulsive systems are typically developed using first-order state-space models, which become impractical for high-order nonlinear impulsive systems due to the increased system dimension and the resulting complexity in controller design. To address this issue, a high-order fully actuated system approach-based adaptive control strategy is developed, enabling direct stabilization without order reduction. An event-triggered mechanism is jointly designed to reduce unnecessary control updates and communication load. The proposed method guarantees stability and boundedness of all closed-loop signals under impulsive effects. Simulation results validate the effectiveness of the approach.

A Differentiable Variational Framework for Joint Optimal Sensor Placement and Physics-Constrained Field Reconstruction

Xiaodong Liu Dalian University of Technology Peoples Rep of China Co-Author(s):    Xiaodong Liu

Abstract: Recent advances have focused on end-to-end differentiable frameworks for jointly optimizing sensor placement and field reconstruction in high-dimensional physical systems. This task can be formulated as a PDE-constrained inverse problem, where latent states are recovered from sparse observations under governing partial differential equations. By embedding physical operators into a differentiable computational graph, sensor locations and model parameters can be optimized simultaneously via backpropagation. Unlike conventional greedy or heuristic methods, such approaches exploit sensitivity information from PDE residuals to identify informative sampling configurations. This leads to sensor placements that improve reconstruction accuracy while preserving physical consistency and enforcing conservation laws. Numerical experiments on transport--diffusion systems demonstrate improved robustness and convergence compared to baseline methods. The use of automatic differentiation for non-convex inverse problems is further examined, along with potential extensions to nonlinear and time-dependent PDE systems.

Data-Driven Distributed Fault-Tolerant Control Based on Stable Kernel Representation and Stable Image Representation

Ying Yang Peking University Peoples Rep of China Co-Author(s):    Shuyu Ding and Ying Yang

Abstract: This paper proposes a data-driven distributed fault-tolerant control (DFTC) method based on the unified stable kernel representation (SKR) and stable image representation (SIR) framework. Firstly, for each multi-input multi-output subsystem, a class of multi-dimensional residual generators is constructed based on data-driven distributed observers. Then, a distributed representation of the interconnected system is established within the SKR and SIR framework. In this representation, the input-output signal space of each subsystem is decomposed into the nominal image subspace and the residual subspace that characterizes the combined effects of faults, uncertainties and coupling dynamics. This decomposition offers a structured representation that integrates monitoring and control perspectives, enabling systematic integration of fault information into controller design. Based on Youla parameterization, a distributed controller reconfiguration scheme is developed, where residual-driven compensators are designed via a model matching formulation. Sufficient conditions for global closed-loop stability are derived using the small-gain theorem. The developed control architecture is fully distributed and allows plug-and-play implementation. Simulation results on a benchmark power system validate the effectiveness of the proposed method in achieving fault compensation and performance recovery.

An Improved Measurement Noise Covariance Estimation Method Based on Envelope Pseudo-measurement System in Adaptive Kalman Filter

Baochang Zhang Beihang University Peoples Rep of China Co-Author(s):    Liuyang Jiang, Xiaolin Zhang, Guohui Zheng, Baochang Zhang

Abstract: Accurate estimation of the measurement noise covariance matrix is crucial for adaptive Kalman filtering in non-stationary environments. Existing methods mainly couple measurement noise covariance matrix estimation with state estimation, while the emerging second-order mutual difference technique achieves state-independent estimation through redundant measurements, but at the cost of requiring additional sensor resources. This paper introduces a novel pseudo-measurement approach that substitutes redundant measurements to mitigate this limitation by using the mean of the upper and lower envelopes of the measurements as the pseudo-measurement. Our analysis reveals that traditional envelope construction methods often suffer from coverage gaps; therefore, we categorize the envelope signals into three trend-based groups and address them comprehensively. Additionally, outliers are detected using the interquartile range (IQR) and replaced by the mean of the neighboring measurements at time steps k-1 and k+1. The performance of the proposed improved envelope pseudo-measurement noise covariance estimate (IEPMNCE) method was evaluated through numerical and motion model simulations. The numerical simulation results show that systematic errors do not affect IEPMNCE. Furthermore, IEPMNCE demonstrates greater resistance to outliers, yielding more accurate state estimation than other methods in motion model simulation. These findings highlight IEPMNCE`s effectiveness in improving the measurement noise covariance matrix estimation accuracy for adaptive filtering and dynamic environments.

Sparse Tensor CCA via Manifold Optimization for Multi-View Learning

Yanjiao Zhu Sun Yat-sen University Peoples Rep of China Co-Author(s):    Yanjiao Zhu, Wanquan Liu, Xianchao Xiu, Jianqin Sun

Abstract: Tensor canonical correlation analysis (TCCA) has garnered significant attention due to its effectiveness in capturing high-order correlations in multi-view learning. However, existing TCCA methods often underemphasize the characterization of individual structures and lack algorithmic convergence guarantees. In order to deal with these challenges, we propose a novel sparse TCCA model called STCCA-L, which integrates sparse regularization of canonical matrices and Laplacian regularization of multi-order graphs into the TCCA framework, thereby effectively exploiting the geometric structure of individual views. To solve this non-convex model, we develop an efficient alternating manifold proximal gradient algorithm based on manifold optimization, which avoids computationally expensive full tensor decomposition and leverages a semi-smooth Newton method for resolving the subproblem. Furthermore, we rigorously prove the convergence of the algorithm and analyze its complexity. Experimental results on eight benchmark datasets demonstrate the superior classification performance of the proposed method. Notably, on the 3Sources dataset, it achieves improvements of at least 4.50\% in accuracy and 6.77\% in F1 score over competitors. Our code is available at https://github.com/zhudafa/STCCA-L.