Yuping Liu

A fundamental challenge in industrial instrumentation is the reliable measurement and monitoring of nonstationary processes, where the sensitivity and accuracy of traditional fault detection systems are frequently jeopardized by changing spatial and temporal variations. This paper introduces the Extended Temporal and Spatial Stationary Subspace Network (ETS³N), a novel framework designed to enhance measurement reliability by jointly modeling temporal homogeneity and spatial heterogeneity in multivariate process signals. In order to maintain the time-space-dependent coherence necessary for local stationarity assumptions, the architecture incorporates three essential units: (1) the Temporal Homogeneity Enumeration Unit (THEU), which reinforces intrinsic temporal coherence to support robust feature extraction; (2) the Spatial Heterogeneity Extraction Unit (SHEU), which constructs a triple-branch convolutional structure with functional independence and diversity constraints to decouple heterogeneous sensor responses; and (3) the Spatio-Temporal Amalgamation Decoupling Unit (STADU), which separates stationary and nonstationary components for interpretable monitoring. With the help of this deep measurement model, latent process dynamics can be accurately extracted, improving fault detection, measurement, and process state comprehension. Validation on both simulated and real-world industrial datasets, such as the blast furnace iron production process and the Secure Water Treatment (SWaT) system, shows that ETS³N outperforms current techniques in terms of sensitivity, stability, and fault isolation. The suggested method aids in the creation of future measurement systems for intricate, dynamic industrial settings.

This paper introduces a novel control strategy for precise motion control of piezoelectric stack actuators (PSAs) under load variations, stochastic disturbances, and model uncertainties. The approach combines fixed-time, nonsingular terminal sliding mode control (FNTSMC) with an adaptive interval type-2 fuzzy neural network (IT2FNN). Specifically, the FNTSMC ensures robust tracking via a fast, fixed-time nonlinear sliding surface and reaching law, while the IT2FNN adaptively estimates and compensates for lumped disturbances, thereby enhancing robustness. Moreover, closed-loop system stability is rigorously proven via Lyapunov theory. Experiments on a PSA test platform under lumped disturbances demonstrate superior tracking performance and robustness compared to existing sliding mode control methods.

Indoor scene recognition is a crucial component in vision-and-language navigation (VLN), which involves guiding an agent to navigate through unseen, photo-realistic environments using natural language instructions. However, existing researches have not yet endowed VLN agents with robust indoor scene recognition capabilities, which are essential for enhancing the agent’s spatial perception and environmental understanding. In this paper, we introduce the Room-to-Room Scene Recognition (R2R-SR) dataset, designed to enhance the spatial perception capabilities of agents in VLN. The R2R-SR dataset includes 7,283 images categorized into 10 room classes, sourced from the Matterport3D (MP3D) simulator. To alleviate the imbalance in training data distribution across room categories, we propose a panoramic environment generation technique using Llava and Stable Diffusion for data Augmentation (LSDA). Our proposed LSDA-based model achieves a Top-1 accuracy of 75.9%, outperforming the best competitors in the convolutional neural network (CNN) and vision transformer (ViT) domains. The qualitative results highlight the model’s proficiency in identifying distinct features across various room categories, showcasing its effectiveness in indoor scene recognition tasks.

Vision-and-Language Navigation (VLN) is a challenging task that requires a robot to navigate in photo-realistic environments with human natural language promptings. Recent studies aim to handle this task by constructing the semantic spatial map representation of the environment, and then leveraging the strong ability of reasoning in large language models for generalizing code for guiding the robot navigation. However, these methods face limitations in instance-level and attribute-level navigation tasks as they cannot distinguish different instances of the same object. To address this challenge, we propose a new method, namely, Instance-aware Visual Language Map (IVLMap), to empower the robot with instance-level and attribute-level semantic mapping, where it is autonomously constructed by fusing the RGBD video data collected from the robot agent with special-designed natural language map indexing in the bird’s-in-eye view. Such indexing is instance-level and attribute-level. In particular, when integrated with a large language model, IVLMap demonstrates the capability to i) transform natural language into navigation targets with instance and attribute information, enabling precise localization, and ii) accomplish zero-shot end-to-end navigation tasks based on natural language commands. Extensive navigation experiments are conducted. Simulation results illustrate that our method can achieve an average improvement of 14.4% in navigation accuracy. Code and demo are released at https://ivlmap.github.io/.

The operational performance of Flexible Hinge Piezoelectric Stages (FHPSs), essential components in precision engineering, is fundamentally constrained by the inherent hysteresis of the piezoelectric actuator (PEA). A significant deficiency in prevailing characterization methods is their failure to consider the dynamic nature of the mechanical preload exerted by the flexible hinge. This position-dependent preload induces substantial deviations in the PEA’s response characteristics, thereby compromising the predictive accuracy of conventional design frameworks. To address this limitation, this paper proposes a Force-Dependent Prandtl–Ishlinskii (FPI) model that explicitly formulates the PEA’s hysteretic behavior as a function of variable preload conditions. The FPI model is subsequently integrated into a comprehensive FPI-FFEM characterization framework. Within this framework, a Force-analyzed Finite Element Method (FFEM) is utilized to compute the dynamic preload throughout the actuator’s operational stroke. This information, notably neglected in conventional FEM analysis, is essential to the fidelity of the proposed FPI model. Experimental validation demonstrates the superior fidelity of the FPI model in comparison to the traditional PI model for tracking preload-induced nonlinearities. Furthermore, the complete FPI-FFEM framework exhibits substantially enhanced prediction accuracy relative to both conventional PI-FEM and advanced LDPI-FEM methodologies, as demonstrated by a significant reduction in the Mean Absolute Error (MAE).