Error Propagation Research Articles

Non-intrusive thermal load disaggregation and forecasting for effective HVAC systems

Non-Intrusive Thermal Load Monitoring (NITLM) tracks the sub-loads generated by each heat source (e.g. occupants, equipment, solar radiation etc.) from the total thermal load and indirectly provides a room’s thermal properties without additional sensors. Since sub-loads can improve the efficiency of HVAC systems, NITLM is a very attractive technology for building energy management. NITLM has traditionally focused on analyzing past and present sub-loads. However, by forecasting future sub-loads, HVAC systems will be able to schedule operations that take into account the thermal properties of future rooms. This work focuses on a new NITLM framework that forecasts future sub-loads based on the current and past total thermal loads. In experiments, we selected occupant loads that are closely connected to HVAC systems and performed sub-load forecasting using two types of approaches. One is a two-step approach that separately performs them in turn. This approach use separately trained model for disaggregation and forecasting, this allow us to fine-tuning the hyper-parameter for dedicate model. Moreover, the two-step approach can take into account the different properties and difficulties of each inference, resulting in smaller errors in sub-load forecasting. The other is an integrated approach. This approach combines load disaggregation and forecasting into a single estimation process, eliminating error propagation and reducing overall error in sub-load forecasting. Moreover, this approach utilizes the Adaptive Moment Estimation (Adam) algorithm for effective parameter optimization, enabling complex training and improving the accuracy of sub-load forecasting. We conducted evaluations of thermal load disaggregation and forecasting across a range of realistic building scenarios. The findings indicate that the integrated approach predicts sub-loads with a MAE that is up to 34.9% lower than that of the two-step approach. Additionally, it identifies occupants presence with an 18.5% higher F-score. This demonstrates its enhanced suitability for accurately predicting sub-loads and estimating future occupancy schedules.

L-PCM: Localization and Point Cloud Registration-Based Method for Pose Calibration of Mobile Robots

The autonomous navigation of mobile robots contains three parts: map building, global localization, and path planning. Precise pose data directly affect the accuracy of global localization. However, the cumulative error problems of sensors and various estimation strategies cause the pose to have a large gap in data accuracy. To address these problems, this paper proposes a pose calibration method based on localization and point cloud registration, which is called L-PCM. Firstly, the method obtains the odometer and IMU (inertial measurement unit) data through the sensors mounted on the mobile robot and uses the UKF (unscented Kalman filter) algorithm to filter and fuse the odometer data and IMU data to obtain the estimated pose of the mobile robot. Secondly, the AMCL (adaptive Monte Carlo localization) is improved by combining the UKF fusion model of the IMU and odometer to obtain the modified global initial pose of the mobile robot. Finally, PL-ICP (point to line-iterative closest point) point cloud registration is used to calibrate the modified global initial pose to obtain the global pose of the mobile robot. Through simulation experiments, it is verified that the UKF fusion algorithm can reduce the influence of cumulative errors and the improved AMCL algorithm can optimize the pose trajectory. The average value of the position error is about 0.0447 m, and the average value of the angle error is stabilized at about 0.0049 degrees. Meanwhile, it has been verified that the L-PCM is significantly better than the existing AMCL algorithm, with a position error of about 0.01726 m and an average angle error of about 0.00302 degrees, effectively improving the accuracy of the pose.

Prediction of Human Reaching Pose Sequences in Human–Robot Collaboration

Abstract In human–robot collaboration, robots and humans must work together in shared, overlapping, workspaces to accomplish tasks. If human and robot motion can be coordinated, then collisions between robot and human can seamlessly be avoided without requiring either of them to stop work. A key part of this coordination is anticipating humans’ future motion so robot motion can be adapted proactively. In this work, a generative neural network predicts a multi-step sequence of human poses for tabletop reaching motions. The multi-step sequence is mapped to a time-series based on a human speed versus motion distance model. The input to the network is the human’s reaching target relative to current pelvis location combined with current human pose. A dataset was generated of human motions to reach various positions on or above the table in front of the human starting from a wide variety of initial human poses. After training the network, experiments showed that the predicted sequences generated by this method matched the actual recordings of human motion within an L2 joint error of 7.6 cm and L2 link roll–pitch–yaw error of 0.301 rad on average. This method predicts motion for an entire reach motion without suffering from the exponential propagation of prediction error that limits the horizon of prior works.

Segmentation and Tracking Based on Equalized Memory Matching Network and Its Application in Electric Substation Inspection

With the wide application of deep learning, power inspection technology has made great progress. However, substation inspection videos often present challenges such as complex backgrounds, uneven lighting distribution, variations in the appearance of power equipment targets, and occlusions, which increase the difficulty of object segmentation and tracking, thereby adversely affecting the accuracy and reliability of power equipment condition monitoring. In this paper, a pixel-level equalized memory matching network (PEMMN) for power intelligent inspection segmentation and tracking is proposed. Firstly, an equalized memory matching network is designed to collect historical information about the target using a memory bank, in which a pixel-level equalized matching method is used to ensure that the reference frame information can be transferred to the current frame reliably, guiding the segmentation tracker to focus on the most informative region in the current frame. Then, to prevent memory explosion and the accumulation of segmentation template errors, a mask quality evaluation module is introduced to obtain the confidence level of the current segmentation result so as to selectively store the frames with high segmentation quality to ensure the reliability of the memory update. Finally, the synthetic feature map generated by the PEMMN and the mask quality assessment strategy are unified into the segmentation tracking framework to achieve accurate segmentation and robust tracking. Experimental results show that the method performs excellently on real substation inspection scenarios and three generalized datasets and has high practical value.

Cascaded framework for earthquake building damage detection combining spatial and frequency domain feature integration

Abstract. Building collapse is a major cause of casualties after an earthquake, so accurately extracting building damage information is critical for post-earthquake assessment and rescue. Currently, most deep learning methods focus on the end-to-end detection of building collapse. However, in real-world earthquake scenarios, the end-to-end computational process often lacks flexibility and struggles to meet the requirements of rapid emergency response. To address this issue, this paper proposes a cascaded framework that combines pre-earthquake building extraction and post-earthquake building damage classification. The proposed framework includes two sections: (1) Progressive building semantic segmentation model in the joint frequency domain. This model is designed to accurately extract buildings prior to an earthquake, with the goal of minimizing error propagation throughout the cascading process. The model addresses the spatial similarity of buildings under complicated backgrounds, as well as the high internal heterogeneity of buildings, by utilizing frequency domain techniques. It compensates for the shortcomings of traditional models in terms of incomplete information extraction through the effective integration of global and local information. Finally, the model employs edge priors for edge regularization. (2) Rapid building damage classification process. Based on the accurate building extraction results, a fast and efficient classification process is developed. This process uses a simple and lightweight classification network to effectively extract building damage information caused by the earthquake. The superiority of the proposed framework is validated through comparison with traditional cascading architectures and end-to-end models. The results show that the cascading framework not only provides accurate pre-earthquake building extraction, but also enables efficient and accurate post-earthquake damage classification, which meets the requirements of rapid post-earthquake emergency response. This balance of accuracy and speed is essential for effective disaster management and recovery.

BCSLinker: automatic method for constructing a knowledge graph of venous thromboembolism based on joint learning

BackgroundVenous thromboembolism (VTE) is characterized by high morbidity, mortality, and complex treatment. A VTE knowledge graph (VTEKG) can effectively integrate VTE-related medical knowledge and offer an intuitive description and analysis of the relations between medical entities. However, current methods for constructing knowledge graphs typically suffer from error propagation and redundant information.MethodsIn this study, we propose a deep learning-based joint extraction model, Biaffine Common-Sequence Self-Attention Linker (BCSLinker), for Chinese electronic medical records to address the issues mentioned above, which often occur when constructing a VTEKG. First, the Biaffine Common-Sequence Self-Attention (BCsSa) module is employed to create global matrices and extract entities and relations simultaneously, mitigating error propagation. Second, the multi-label cross-entropy loss is utilized to diminish the impact of redundant information and enhance information extraction.ResultsWe used the electronic medical record data of VTE patients from a tertiary hospital, achieving an F1 score of 86.9% on BCSLinker. It outperforms the other joint entity and relation extraction models discussed in this study. In addition, we developed a question-answering system based on the VTEKG as a structured data source.ConclusionThis study has constructed a more accurate and comprehensive VTEKG that can provide reference for diagnosing, evaluating, and treating VTE as well as supporting patient self-care, which is of considerable clinical value.

Learning content-aware feature fusion for guided depth map super-resolution

RGB-D data including paired RGB color images and depth maps is widely used in downstream computer vision tasks. However, compared with the acquisition of high-resolution color images, the depth maps captured by consumer-level sensors are always in low resolution. Within decades of research, the most state-of-the-art (SOTA) methods of depth map super-resolution cannot adaptively tune the guidance fusion for all feature positions by channel-wise feature concatenation with spatially sharing convolutional kernels. This paper proposes JTFNet to resolve this issue, which simulates the traditional Joint Trilateral Filter (JTF). Specifically, a novel JTF block is introduced to adaptively tune the fusion pattern between the color features and the depth features for all feature positions. Moreover, based on the variant of JTF block whose target features and guidance features are in the cross-scale shape, the fusion for depth features is performed in a bi-directional way. Therefore, the error accumulation along scales can be effectively mitigated by iteratively HR feature guidance. Compared with the SOTA methods, the sufficient experiment is conducted on the mainstream synthetic datasets and real datasets, i.e., Middlebury, NYU and ToF-Mark, which shows remarkable improvement of our JTFNet.

Direct Determination of Absolute Radical Quantum Yields in Hydroxyl and Sulfate Radical-Based Treatment Processes

Direct Determination of Absolute Radical Quantum Yields in Hydroxyl and Sulfate Radical-Based Treatment Processes

Large-scale DSM registration via motion averaging

Abstract. Generating wide-area digital surface models (DSMs) requires registering a large number of individual, and partially overlapped DSMs. This presents a challenging problem for a typical registration algorithm, since when a large number of observations from these multiple DSMs are considered, it may easily cause memory overflow. Sequential registration algorithms, although can significantly reduce the computation, are especially vulnerable for small overlapped pairs, leading to a large error accumulation. In this work, we propose a novel solution that builds the DSM registration task as a motion averaging problem: pair-wise DSMs are registered to build a scene graph, with edges representing relative poses between DSMs. Specifically, based on the grid structure of the large DSM, the pair-wise registration is performed using a novel nearest neighbor search method. We show that the scene graph can be optimized via an extremely fast motion average algorithm with O(N) complexity (N refers to the number of images). Evaluation of high-resolution satellite-derived DSM demonstrates significant improvement in computation and accuracy.

High-Order Randomized Compiler for Hamiltonian Simulation

Hamiltonian simulation is known to be one of the fundamental building blocks of a variety of quantum algorithms such as its most immediate application, that of simulating many-body systems to extract their physical properties. In this work, we present qSWIFT, a high-order randomized algorithm for Hamiltonian simulation. In qSWIFT, the required number of gates for a given precision is independent of the number of terms in the Hamiltonian, while the systematic error is exponentially reduced with regard to the order parameter. In this respect, our qSWIFT is a higher-order counterpart of the previously proposed quantum stochastic drift protocol (qDRIFT), the number of gates in which scales linearly with the inverse of the precision required. We construct the qSWIFT channel and establish a rigorous bound for the systematic error quantified by the diamond norm. qSWIFT provides an algorithm to estimate given physical quantities by using a system with one ancilla qubit, which is as simple as other product-formula-based approaches such as regular Trotter-Suzuki decompositions and qDRIFT. Our numerical experiment reveals that the required number of gates in qSWIFT is significantly reduced compared to qDRIFT. In particular, the advantage is significant for problems where high precision is required; e.g., to achieve a systematic relative propagation error of 10−6, the required number of gates in third-order qSWIFT is 1000 times smaller than that of qDRIFT. Published by the American Physical Society 2024

JEEMRC: Joint Event Detection and Extraction via an End-to-End Machine Reading Comprehension Model

Event extraction (EE) generally contains two subtasks: viz., event detection and argument extraction. Owing to the success of machine reading comprehension (MRC), some researchers formulate EE into MRC frameworks. However, existing MRC-based EE techniques are pipeline methods that suffer from error propagation. Moreover, the correlation between event types and argument roles is pre-defined by experts, which is time-consuming and inflexible. To avoid these issues, event detection and argument extraction are formalized as joint MRC. Different from previous methods, which just generate questions for argument roles for identified event types, questions are generated for all arguments that appear in the given sentence in our approach. Moreover, an end-to-end MRC model, JEEMRC, is proposed, which consists of an event classifier and a machine reader with a coarse-to-fine attention mechanism. Our proposed model can train two subtasks jointly to alleviate error propagation and utilizes interaction information between event types and argument roles to improve the performance of both tasks. Experiments on ACE 2005 verify that our JEEMRC achieves competitive results compared with previous work. In addition, it performs well when detecting events and extracting arguments in data-scarce scenarios.

A novel deep learning rainfall–runoff model based on Transformer combined with base flow separation

ABSTRACT Precise long-term runoff prediction holds crucial significance in water resource management. Although the long short-term memory (LSTM) model is widely adopted for long-term runoff prediction, they encounter challenges such as error accumulation and low computational efficiency. To address these challenges, we utilized a novel method to predict runoff based on a Transformer and the base flow separation approach (BS-Former) in the Ningxia section of the Yellow River Basin. To evaluate the effectiveness of the Transformer model and its responsiveness to the base flow separation technique, we constructed LSTM and artificial neural network (ANN) models as benchmarks for comparison. The results show that Transformer outperforms the other models in terms of predictive performance and that base flow separation significantly improves the performance of the Transformer model. Specifically, the performance of BS-Former in predicting runoff 7 days in advance is comparable to that of the BS-LSTM and BS-ANN models with lead times of 4 and 2 days, respectively. In general, the BS-Former model is a promising tool for long-term runoff prediction.

A Parallel Model for Jointly Extracting Entities and Relations

Extracting relational triples from a piece of text is an essential task in knowledge graph construction. However, most existing methods either identify entities before predicting their relations, or detect relations before recognizing associated entities. This order may lead to error accumulation because once there is an error in the initial step, it will accumulate to subsequent steps. To solve this problem, we propose a parallel model for jointly extracting entities and relations, called PRE-Span, which consists of two mutually independent submodules. Specifically, candidate entities and relations are first generated by enumerating token sequences in sentences. Then, two independent submodules (Entity Extraction Module and Relation Detection Module) are designed to predict entities and relations. Finally, the predicted results of the two submodules are analyzed to select entities and relations, which are jointly decoded to obtain relational triples. The advantage of this method is that all triples can be extracted in just one step. Extensive experiments on the WebNLG*, NYT*, NYT and WebNLG datasets show that our model outperforms other baselines at 94.4%, 88.3%, 86.5% and 83.0%, respectively.

Bi-scale car-following model calibration based on corridor-level trajectory

The precise estimation of macroscopic traffic parameters, such as travel time and fuel consumption, is essential for the optimization of traffic management systems. Despite its importance, the comprehensive acquisition of vehicle trajectory data for the calculation of these macroscopic measures presents a challenge. To bridge this gap, this study aims to calibrate car-following models capable of predicting both microscopic measures and macroscopic measures. We conduct a numerical analysis to trace the cumulative process of model prediction errors across various measurements, and our findings indicate that macroscopic measures encapsulate the accumulation of model errors. By incorporating macroscopic measures into vehicle model calibration, we can mitigate the impact of noise on microscopic data measurements. We compare three car-following model calibration methods: MiC (using microscopic measurements), MaC (using macroscopic measurements), and BiC (using both microscopic and macroscopic measurements)—utilizing real-world trajectory data. The BiC method emerges as the most successful in reconstructing vehicle trajectories and accurately estimating travel time and fuel consumption, whereas the MiC method leads to overfitting and inaccurate macro-measurement predictions. This study underscores the importance of bi-scale calibration for precise traffic and energy consumption predictions, laying the groundwork for future research aimed at enhancing traffic management strategies.

A dynamic neural resource model bridges sensory and working memory

Probing memory of a complex visual image within a few hundred milliseconds after its disappearance reveals significantly greater fidelity of recall than if the probe is delayed by as little as a second. Classically interpreted, the former taps into a detailed but rapidly decaying visual sensory or ‘iconic’ memory (IM), while the latter relies on capacity-limited but comparatively stable visual working memory (VWM). While iconic decay and VWM capacity have been extensively studied independently, currently no single framework quantitatively accounts for the dynamics of memory fidelity over these time scales. Here, we extend a stationary neural population model of VWM with a temporal dimension, incorporating rapid sensory-driven accumulation of activity encoding each visual feature in memory, and a slower accumulation of internal error that causes memorized features to randomly drift over time. Instead of facilitating read-out from an independent sensory store, an early cue benefits recall by lifting the effective limit on VWM signal strength imposed when multiple items compete for representation, allowing memory for the cued item to be supplemented with information from the decaying sensory trace. Empirical measurements of human recall dynamics validate these predictions while excluding alternative model architectures. A key conclusion is that differences in capacity classically thought to distinguish IM and VWM are in fact contingent upon a single resource-limited WM store.

A dynamic neural resource model bridges sensory and working memory.

Probing memory of a complex visual image within a few hundred milliseconds after its disappearance reveals significantly greater fidelity of recall than if the probe is delayed by as little as a second. Classically interpreted, the former taps into a detailed but rapidly decaying visual sensory or 'iconic' memory (IM), while the latter relies on capacity-limited but comparatively stable visual working memory (VWM). While iconic decay and VWM capacity have been extensively studied independently, currently no single framework quantitatively accounts for the dynamics of memory fidelity over these time scales. Here, we extend a stationary neural population model of VWM with a temporal dimension, incorporating rapid sensory-driven accumulation of activity encoding each visual feature in memory, and a slower accumulation of internal error that causes memorized features to randomly drift over time. Instead of facilitating read-out from an independent sensory store, an early cue benefits recall by lifting the effective limit on VWM signal strength imposed when multiple items compete for representation, allowing memory for the cued item to be supplemented with information from the decaying sensory trace. Empirical measurements of human recall dynamics validate these predictions while excluding alternative model architectures. A key conclusion is that differences in capacity classically thought to distinguish IM and VWM are in fact contingent upon a single resource-limited WM store.

An enhanced motorcycle tyre model characterised through experimental riding data

Tyre-road interaction governs motorcycle dynamics; however, the most widespread tyre model formulations must be characterised through a dedicated test bench on the lab or road, unavailable to many interested subjects. This article proposed a new tyre model formulation, conceived to be characterised through riding data using standard instrumentation. Albeit its coefficients are identified through quasi-static, uncombined slip manoeuvres, the model addresses transient, combined manoeuvres and is adaptive to road friction levels and static weight through statistical relationships from the literature. A pre-existing formulation was improved and expanded. The model’s behaviour in demanding conditions was investigated through a high-fidelity simulation environment, using a Magic Formula tyre model as the reference. Next, the characterisation procedure was carried out using actual riding data. The model’s accuracy is shown by reproducing numerically one of the manoeuvres and through comparison with the results of a bench test. The proposed model could correctly reproduce the primary behaviour of a Magic Formula model, also concerning tyre moments and steering torque. Characterising the tyre model through real riding data proved feasible, and its robust formulation limited the propagation of estimation errors. The proposed tyre model formulation and characterisation procedure should interest, among others, those subjects that lack access to a tyre testing machine.

Take a prior from other tasks for severe blur removal

Recovering clear structures from severely blurry inputs is a huge challenge due to the detail loss and ambiguous semantics. Although segmentation maps can help deblur facial images, their effectiveness is limited in complex natural scenes because they ignore the detailed structures necessary for deblurring. Furthermore, direct segmentation of blurry images may introduce error propagation. To alleviate the semantic confusion and avoid error propagation, we propose utilizing high-level vision tasks, such as classification, to learn a comprehensive prior for severe blur removal. We propose a feature learning strategy based on knowledge distillation, which aims to learn the priors with global contexts and sharp local structures. To integrate the priors effectively, we propose a semantic prior embedding layer with multi-level aggregation and semantic attention. We validate our method on natural image deblurring benchmarks by introducing the priors to various models, including UNet and mainstream deblurring baselines, to demonstrate its effectiveness and generalization ability. The results show that our approach outperforms existing methods on severe blur removal with our plug-and-play semantic priors.

Cumulative error correction of inertial navigation systems using LIDAR sensors and extended Kalman filter

Autonomous robots have gained significant attention in research due to their ability to facilitate human work. Navigation systems, particularly localization, present a challenge in autonomous robots. The inertial navigation system is a localization system that uses inertial sensors and a wheel odometer to estimate the robot’s relative position to the initial position. However, the system is susceptible to continuous error accumulation over time due to factors like sensor noise and wheel slip. To address these issues, external sensors are required to measure the robot’s position in the environment. The extended Kalman filter (EKF) method is utilized to estimate the robot’s position based on wheel odometer and light detection and ranging (LIDAR) sensor measurements. In the prediction stage, the input to the EKF is the position measurement from the wheel odometer, while the LIDAR sensor’s position measurement is used in the update stage to improve the prediction stage results. The test results reveal that the EKF’s estimated position has a lower average error compared to the position measurement using the wheel odometer. Therefore, it can be concluded that the EKF technique is effectively applied to the robot and can correct the wheel odometer's cumulative error with the assistance of the LIDAR sensor.

On-chip ultra-high rejection and narrow bandwidth filter based on coherency-broken cascaded cladding-modulated gratings

Bragg filters are of essential importance for chip-scale photonic systems. However, the implementation of filters with sub-nanometer bandwidth and rejection beyond 70 dB is hindered by the high index contrast of the silicon-on-insulator platform, which makes filters prone to fabrication imperfections. In this paper, we propose to combine coherency-broken cascading architecture and cladding modulation to circumvent the intrinsic limitation. The cascading architecture effectively prevents the accumulation of phase errors, while the cladding modulation offers additional design freedom to reduce the coupling coefficient. A bimodal Bragg filter with a testing-equipment-limited rejection level of 74 dB and a 40 dB bandwidth of 0.44 nm is experimentally demonstrated. The minimum feature size is 90 nm, which significantly relieves the fabrication constraints.