Particle Filtering Technique Research Articles

Model uncertainty quantification of a degradation model of miter gates using normalizing flow-based likelihood-free inference

Degradation modeling is essential for failure prognostics of miter gates and subsequent risk-informed maintenance or operational decision-making. Typically, degradation models are formulated based on certain assumptions and simplifications and may not fully capture the complexity of underlying degradation mechanisms. Even minor errors in such models may lead to significant discrepancies in failure prognostics due to error accumulation over time. Aiming to improve the accuracy of the degradation model and ensure reliable failure prognostics, this article proposes a degradation model updating framework for miter gates using normalizing flow-based likelihood-free inference, accounting for both model parameter uncertainty and model form uncertainty (i.e., model bias). The proposed work consists of two phases: the offline training phase and the online updating phase. During the offline training phase, the unknown model bias term is modeled as a random noise with uncertain standard deviation. Synthetic data are then generated using a physical model based on prior knowledge of the uncertain model parameters and model bias. The synthetic data are utilized to train an inference model, which has a summary network for data compression and a conditional invertible neural network for parameter estimation. In the online updating phase, the trained inference model uses strain observations to obtain posterior samples. The Gaussian mixture model and dual particle filter techniques are utilized to recursively update posterior samples, thereby improving the estimation accuracy. Subsequently, model bias is inversely estimated using the degradation model, which uses the updated model parameters and the gap length from the inference model. Following that, a regression model is constructed to correct the biases inherent in the simplified degradation model. This process is implemented iteratively over time to perform continuous model updating and failure prognostics. Results of a case study demonstrate the efficacy of the proposed framework in reducing both model parameter uncertainty and recovering model biases and thereby improving the accuracy of failure prognostics of miter gates.

Improved Particle Filter Algorithm for Multi-Target Detection and Tracking.

In modern radar detection systems, the particle filter technique has become one of the core algorithms for real-time target detection and tracking due to its good nonlinear and non-Gaussian system state estimation capability. However, when dealing with complex dynamic scenes, the traditional particle filter algorithm exposes obvious deficiencies. The main expression is that the sample degradation is serious, which leads to a decrease in estimation accuracy. In multi-target states, the algorithm is difficult to effectively distinguish and stably track each target, which increases the difficulty of state estimation. These problems limit the application potential of particle filter technology in multi-target complex environments, and there is an urgent need to develop a more advanced algorithmic framework to enhance its robustness and accuracy in complex scenes. Therefore, this paper proposes an improved particle filter algorithm for multi-target detection and tracking. Firstly, the particles are divided into tracking particles and searching particles. The tracking particles are used to maintain and update the trajectory information of the target, and the searching particles are used to identify and screen out multiple potential targets in the environment, to sufficiently improve the diversity of the particles. Secondly, the density-based spatial clustering of applications with noise is integrated into the resampling phase to improve the efficiency and accuracy of particle replication, so that the algorithm can effectively track multiple targets. Experimental result shows that the proposed algorithm can effectively improve the detection probability, and it has a lower root mean square error (RMSE) and a stronger adaptability to multi-target situation.

A particle filter-based approach for real-time temperature estimation in a lithium-ion battery module during the cooling-down process

Temperature exerts a remarkable influence on the safety, performance, and lifespan of lithium-ion (Li-ion) batteries. Consequently, the real-time monitoring of this variable within battery cells emerges as both crucial and a significant challenge for battery thermal management systems. This paper introduces an approach for real-time temperature estimation in a Li-ion battery module during the cooling-down process by applying the particle filter (PF) technique. A battery module comprising fifteen cylindrical lithium cobalt oxide cells connected in series was employed for experimentation. The battery module was arranged into three rows, the PF employs real-time temperature measurements from the first cell of each string, based on a fractal-time model and an artificial evolution approach for parameter estimation, to estimate two unknown parameters and the temperature of the remaining twelve cells. Furthermore, the temperature measured on the surface of each cell in the module was compared both the PF and computational fluid dynamics (CFD) approach. The results exhibit a favorable agreement between the cell temperature estimates derived from the PF, the CFD simulations, and their respective experimental measurements. Finally, the proposed approach facilitates real-time temperature monitoring in battery modules with a reduced number of temperature sensors, thereby implying a cost reduction in the data acquisition system.

A Robust Evolutionary Particle Filter Technique for Integrated Navigation in Urban Environments via GNSS and 5G Signals

A Robust Evolutionary Particle Filter Technique for Integrated Navigation in Urban Environments via GNSS and 5G Signals

Kernel-Based Particle Filtering for Scalable Inference in Partially Observed Boolean Dynamical Systems

This paper addresses the inference challenges associated with a class of hidden Markov models with binary state variables, known as partially observed Boolean dynamical systems (POBDS). POBDS have demonstrated remarkable success in modeling the ON and OFF dynamics of genes, microbes, and bacteria in systems biology, as well as in network security to represent the propagation of attacks among interconnected elements. Despite existing optimal and approximate inference solutions for POBDS, scalability remains a significant issue due to the computational cost associated with likelihood evaluations and the exploration of extensive parameter spaces. To overcome these challenges, this paper proposes a kernel-based particle filtering approach for large-scale inference of POBDS. Our method employs a Gaussian process (GP) to efficiently represent the expensive-to-evaluate likelihood function across the parameter space. The likelihood evaluation is approximated using a particle filtering technique, enabling the GP to account for various sources of uncertainty, including limited likelihood evaluations. Leveraging the GP’s predictive behavior, a Bayesian optimization strategy is derived for effectively seeking parameters yielding the highest likelihood, minimizing the overall computational burden while balancing exploration and exploitation. The proposed method’s performance is demonstrated using two biological networks: the mammalian cell-cycle network and the T-cell large granular lymphocyte leukemia network.

PSINDy: Probabilistic sparse identification of nonlinear dynamics for temporal process modeling and fault detection

BackgroundMultivariate statistical process monitoring (MSPM) methods have been widely studied and applied for dynamic process monitoring. However, limitations exist in the extant literature, as numerous extant MSPM approaches fail to concurrently account for the governing dynamics and inherent non-linearity of systems. MethodsTo handle the above-mentioned issue, a novel probabilistic dynamic process monitoring algorithm named probabilistic sparse identification of nonlinear dynamics (PSINDy) is proposed. In this algorithm, the Expectation–Maximization (EM) method is employed to estimate the parameters. In the E-step, the particle filtering technique is adopted to calculate corresponding latent expectations whose posterior distributions are not Gaussian. Furthermore, Hotelling's T2 and two novel monitoring statistics, termed Mahalanobis-based predictive error (MPE) and dynamic predictive error (DPE), are designed and utilized for fault detection. Significant FindingsThe effectiveness of the proposed method is validated through the Tennessee Eastman (TE) chemical process and the three-phase flow facility (TPFF).

Deep learning-based activity-aware 3D human motion trajectory prediction in construction

Predicting human motion is a critical requirement in various applications, with particular significance in the construction sector. This task presents significant challenges due to the diverse nature of human actions and the complexities of converting 2D image coordinates to real-world space. In response to these challenges, this paper introduces an innovative deep learning-driven approach to forecast human motion trajectories, with a novel emphasis on activity recognition to improve predictive accuracy. The method utilizes a multi-camera system to extract 2D joint locations, which are then fused using a particle filter technique for 3D pose generation. Using 3D data, deep learning models are developed to first recognize the activity class, and then take it as auxiliary information for predicting the motion trajectory. Through a comprehensive experiment, we evaluated the proposed methodology. While the main innovation of the proposed approach lies in the incorporation of deep learning-based activity recognition into the trajectory prediction system, the experiment results revealed the activity-aware system’s capability to enhance prediction performance by a minimum of 6.4% and up to 16.6% in short-term forecasts, in compare with a conventional approach. Additionally, we analyzed the effects of varying time windows and joint selections on predictive outcomes across diverse scenarios and discussed the implications of these findings. By enhancing the prediction of human motion, this approach holds promise in improving workspace safety while encouraging effective interactions within complex environments.

An outlier-robust Rao–Blackwellized particle filter for underwater terrain-aided navigation

Terrain-aided navigation (TAN) is one of the key techniques for autonomous navigation and positioning of AUVs. This paper develops a novel integrated navigation system that combines the inertial navigation system (INS) and TAN into one filter, in contrast to the conventional INS/TAN model. This method not only provides a position estimate for the carrier but also the INS error, which increases the filtering dimension significantly. Thus, the Rao-Blackwellized particle filter (RBPF) technique is introduced to address the “dimensional disaster” problem. A three-dimensional (3-D) model is established to estimate the horizontal position while compensating for the tidal inaccuracy by extending the tidal depth bias to the state space. Furthermore, the variable and complex underwater environment can potentially result in outlier interference in depth measurements, which significantly degrades the performance of the RBPF-based estimation algorithm. This paper thus presents a 3-D adaptive RBPF method based on the maximum correntropy criterion (MCC) to suppress the influence of measurement outliers on positioning precision. Simulations and experimental tests verify the effectiveness of the proposed algorithm. The results demonstrated that the 3-D model-based AMCCRBPF method can effectively improve robustness to outliers and provide an accurate estimate of the tidal depth bias.

A Physics‐Aware Machine Learning‐Based Framework for Minimizing Prediction Uncertainty of Hydrological Models

AbstractModeling hydrological processes for managing the available water resources effectively is often complex due to the existence of high nonlinearity, and the associated prediction uncertainty mainly arising from model inputs, parameters, and structure. Despite several attempts to quantify the model prediction uncertainty, reducing the same for improving the reliability of models is indispensable for their wider acceptance. This paper presents a novel modeling framework for minimizing the prediction uncertainty in the streamflow simulation of the conceptual hydrological model (HBV) by integrating with the Bayesian‐based Particle Filter technique (PF) and machine learning algorithm (Random Forest algorithm, RF). Initially, the streamflow prediction interval (PI) is derived from the stochastically estimated parameters of the HBV model through the PF technique (HBV‐PF model). As the HBV‐PF model quantifies only parametric uncertainty, the RF algorithm was employed (HBV‐PF‐RF model) for further minimizing the prediction uncertainty by inherently taking care of different sources of uncertainty. The RF algorithm inherently combines the physics of the hydrological system (i.e., process‐based variables) with machine learning‐based approach to minimize the overall prediction uncertainty. The proposed framework was analyzed on Nepal and India's Sunkoshi and Beas River basins, through several statistical performance indices for assessing the accuracy and uncertainty of the model prediction. The framework was observed to be consistently improving the model performance minimizing the uncertainty in both watersheds. Therefore, the proposed framework can be considered to be more reliable in improving the prediction capability of hydrological models.

Generation of Non-Linear Technique Based 6 Hourly Wind Reanalysis Products Using SCATSAT-1 and Numerical Weather Prediction Model Outputs

We combined observations of ocean surface winds from Indian SCATterometer SATellite-1 (SCATSAT-1) with a background wind field from a numerical weather prediction (NWP) model available at National Centre for Medium-Range Weather Forecast (NCMRWF) to generate a 6-hourly gridded hybrid wind product. A distinctive feature of the study is to produce a global gridded wind field from SCATSAT-1 scatterometer passes with spatio-temporal data gaps at regular synoptic hours relevant for forcing models and other NWP studies. We are following the concept from the modern particle filter technique, which does not represent the model probability density function (PDF) as Gaussian. We generated the 6-hourly hybrid winds for 2018 and validated them using the wind speed from daily gridded level-4 SCATSAT-1 winds (L4AW), Cross Calibrated Multi-Platform (CCMP) dataset and global buoy data from National Data Buoy Centre (NDBC). The results suggest the potential of the technique to produce scatterometer winds at the desired temporal frequency with significantly less noise and bias along the swath. The study shows that the generated hybrid winds are of prime quality compared with the already existing daily products available from Indian Space Research Organization (ISRO).

Estimation of Subsea Line Structure Behavior Based on Sequential Data Assimilation with Distributed Sensing

Abstract In this paper, a method that estimates the real-time behavior of subsea line structures based on sequential data assimilation with distributed strain sensors is proposed. A finite element method is used to represent the behavior of subsea line structures and generates ensemble forecasts regarding unknown parameters. A merging particle filter technique is applied to integrate the observation data with the numerical models to calculate the posterior probability density function. The effectiveness of the proposed method is examined through twin experiments. The presented results validate the proposed method's capability to estimate the current state as well as unknown parameters of subsea line structures. The results suggest the advantage of distributed sensors against pointwise sensing when applied to line structures.

Multivariate assimilation of satellite-based leaf area index and ground-based river streamflow for hydrological modelling of irrigated watersheds using SWAT+

• Performance of the SWAT + model with and without data assimilation (DA) is assessed. • Some modules of SWAT + source code are modified to be connected to the DA routine. • Multivariate and univariate assimilations of discharge and LAI data are compared. • High-resolution crop mapping is essential for LAI assimilation in irrigated watersheds. • Joint assimilation of streamflow and Sentinel-2′s LAI results in better simulation of streamflow, vegetation, and irrigation. Vegetation dynamics have different effects on the terrestrial water cycle and therefore play an important role in catchment-scale biophysical and hydrological modeling. Meanwhile, validation of hydrological models of irrigated watersheds concerning vegetation dynamics has rarely been studied. In this study, we propose a combinatorial approach for modelling irrigated watersheds; at first, based on Sentinel-2 (S2), we provide crop/land-use mapping; then we retrieve S2-based leaf area index (LAI) through the neural network algorithm and the PROSAIL model; finally, we employ sequential particle filter (PF) technique to explore the benefits of jointly assimilating high-resolution S2-LAI as well as in-situ river discharge observations for improving the predictive accuracy of SWAT + model. Moreover, we explain how SWAT + source code must be modified to implement the assimilation procedure. The developed methodology is applied to an irrigated watershed in Iran; we provide 66 raster LAI maps with a spatial resolution of 20 m for the study area related to January to December of 2019 as well as crop/land-use mapping of 2019 with a spatial resolution of 10 m. Results show that improvements in the hydrological simulation of LAI, evapotranspiration, and river discharge are achieved when we apply multivariate assimilation of S2-LAI and streamflow observations, compared to univariate assimilation scenarios. Results also reveal that LAI assimilation has a significant influence on the estimation of irrigation volume and timing.

Combining Crop Modeling with Remote Sensing Data Using a Particle Filtering Technique to Produce Real-Time Forecasts of Winter Wheat Yields under Uncertain Boundary Conditions

Within-season crop yield forecasting at national and regional levels is crucial to ensure food security. Yet, forecasting is a challenge because of incomplete knowledge about the heterogeneity of factors determining crop growth, above all management and cultivars. This motivates us to propose a method for early forecasting of winter wheat yields in low-information systems regarding crop management and cultivars, and uncertain weather condition. The study was performed in two contrasting regions in southwest Germany, Kraichgau and Swabian Jura. We used in-season green leaf area index (LAI) as a proxy for end-of-season grain yield. We applied PILOTE, a simple and computationally inexpensive semi-empirical radiative transfer model to produce yield forecasts and assimilated LAI data measured in-situ and sensed by satellites (Landsat and Sentinel-2). To assimilate the LAI data into the PILOTE model, we used the particle filtering method. Both weather and sowing data were treated as random variables, acknowledging principal sources of uncertainties to yield forecasting. As such, we used the stochastic weather generator MarkSim® GCM to produce an ensemble of uncertain meteorological boundary conditions until the end of the season. Sowing dates were assumed normally distributed. To evaluate the performance of the data assimilation scheme, we set up the PILOTE model without data assimilation, treating weather data and sowing dates as random variables (baseline Monte Carlo simulation). Data assimilation increased the accuracy and precision of LAI simulation. Increasing the number of assimilation times decreased the mean absolute error (MAE) of LAI prediction from satellite data by ~1 to 0.2 m2/m2. Yield prediction was improved by data assimilation as compared to the baseline Monte Carlo simulation in both regions. Yield prediction by assimilating satellite-derived LAI showed similar statistics as assimilating the LAI data measured in-situ. The error in yield prediction by assimilating satellite-derived LAI was 7% in Kraichgau and 4% in Swabian Jura, whereas the yield prediction error by Monte Carlo simulation was 10 percent in both regions. Overall, we conclude that assimilating even noisy LAI data before anthesis substantially improves forecasting of winter wheat grain yield by reducing prediction errors caused by uncertainties in weather data, incomplete knowledge about management, and model calibration uncertainty.

The Challenges of Simulating SWE Beneath Forest Canopies are Reduced by Data Assimilation of Snow Depth

AbstractIntermittent snow depth observations can be leveraged with data assimilation (DA) to improve model estimates of snow water equivalent (SWE) at the point scale. A key consideration for scaling assimilation to the basin scale is its performance at forested locations—where canopy‐snow interactions affect snow accumulation and melt, yet are difficult to model and parameterize. We implement a particle filter (PF) technique to assimilate intermittent snow depth observations into the Flexible Snow Model, and validate model outputs against snow density and SWE measurements across adjacent forest and open sites, at two locations with different climates and forest structures. Assimilation reduces snow depth error by 70%–90%, density error by 5%–30%, and SWE error by 50%–70% at forest locations (relative to controls). The PF simulates the seasonal evolution of the snowpack under forest canopy, including where interception losses decrease SWE in the forest during accumulation, and shading reduces melt during the ablation season (relative to open sites). Model outputs are sensitive to canopy‐related parameters, but DA reduces the range in snow depth and SWE estimates resulting from variations or uncertainties in these parameters by over 50%. These results demonstrate that the importance of accurately measuring, estimating, or calibrating canopy‐related parameters is reduced when snow depth observations are assimilated. Finally, we assimilate remotely sensed snow depth observations at 50 m resolution across the East River Basin, Colorado (1,000 km2). Across elevations, the PF increases estimated precipitation at forested sites by ∼15% relative to open sites, likely compensating for excessive sublimation of intercepted snow.

An enhanced particle filter technology for battery system state estimation and RUL prediction

The particle filter (PF) technique can model nonlinear degradation features of battery’s system, and conduct battery state estimation based on noisy measurements. However, PF has some limitations in system state estimation related to sample degeneracy and impoverishment. In addition, its posterior probability density function cannot be updated during the prognostic period due to the absence of new battery measurements. In this work, an enhanced PF technology is proposed to deal with these problems so as to improve PF modeling accuracy for battery state-of-health monitoring and remaining useful life (RUL) prediction. Specifically, an enhanced particles method is proposed to reduce the impact of sample degeneracy and impoverishment in state estimation. An evolving fuzzy predictor is adopted and fused into the enhanced PF structure to deal with the lack of new battery measurements during the prognostic period. The effectiveness of the proposed enhanced PF technology is validated through simulation tests.

Particle Filter Based Range Search Approach for Localization of Radioactive Materials

Wireless Sensor Networks (WSNs) have recently become crucial in monitoring operations. The development of a Data Fusion Algorithm for radioactive source localization utilizing WSN based on the Particle Filter (PF) technique is presented. The localization of an unknown-intensity point radioactive source using sensor nodes measured intensities in Count per Minute (CPM) is considered. The surveillance area is covered by several sensors (radiation detectors) n. Instead of using four sensors, as described in previous research, two consecutive sensors are used in sequence (S1, S2), (S2, S3)……, (Sn-1, Sn) till reaching the last sensor available Sn. Apollonius circle calculated range guides a particle filter for estimating the source location using actual measurements. Compared with other approaches such as the Iterative Pruning Clustering algorithm, more accurate estimates in terms of the error between the estimated source position and ground truth are obtained. The comparison is conducted using the same real measurements data. The Particle Filter based algorithm is implemented in a Xilinx FPGA chip. The architecture is a two sequential steps implementation, where particle generation, weight calculation, and normalization are carried out in parallel during the first step, followed by a sequential or parallelized resampling in the second step. This architecture targets a balance between hardware resources and speed of operation. The future work plan includes security-related studies and complete WSN implementation using μC/FPGA devices.

A remaining useful life estimation method for solenoid valve based on mmWave radar and auxiliary particle filter technique

A remaining useful life estimation method for solenoid valve based on mmWave radar and auxiliary particle filter technique

Iterative identification methods for a class of bilinear systems by using the particle filtering technique

SummaryThis article mainly studies the iterative parameter estimation problems of a class of nonlinear systems. Based on the auxiliary model identification idea, this article utilizes the estimated parameters to construct an auxiliary model, and uses its outputs to replace the unknown noise‐free process outputs, and develops an auxiliary model least squares‐based iterative (AM‐LSI) identification algorithm. For further improving the parameter estimation accuracy, we use a particle filter to estimate the unknown noise‐free process outputs, and derive a particle filtering least squares‐based iterative (PF‐LSI) identification algorithm. During each iteration, the AM‐LSI and PF‐LSI algorithms can make full use of the measured input–output data. The simulation results indicate that the proposed algorithms are effective for identifying the nonlinear systems, and can generate more accurate parameter estimates than the auxiliary model‐based recursive least squares algorithm.

Irrigation Amounts and Timing Retrieval through Data Assimilation of Surface Soil Moisture into the FAO-56 Approach in the South Mediterranean Region

Agricultural water use represents more than 70% of the world’s freshwater through irrigation water inputs that are poorly known at the field scale. Irrigation monitoring is thus an important issue for optimizing water use in particular with regards to the water scarcity that the semi-arid regions are already facing. In this context, the aim of this study is to develop and evaluate a new approach to predict seasonal to daily irrigation timing and amounts at the field scale. The method is based on surface soil moisture (SSM) data assimilated into a simple land surface (FAO-56) model through a particle filter technique based on an ensemble of irrigation scenarios. The approach is implemented in three steps. First, synthetic experiments are designed to assess the impact of the frequency of observation, the errors on SSM and the a priori constraints on the irrigation scenarios for different irrigation techniques (flooding and drip). In a second step, the method is evaluated using in situ SSM measurements with different revisit times (3, 6 and 12 days) to mimic the available SSM product derived from remote sensing observation. Finally, SSM estimates from Sentinel-1 are used. Data are collected on different wheat fields grown in Morocco, for both flood and drip irrigation techniques in addition to rainfed fields used for an indirect evaluation of the method performance. Using in situ data, accurate results are obtained. With an observation every 6 days to mimic the Sentinel-1 revisit time, the seasonal amounts are retrieved with R > 0.98, RMSE < 32 mm and bias < 2.5 mm. Likewise, a good agreement is observed at the daily scale for flood irrigation as more than 70% of the detected irrigation events have a time difference from actual irrigation events shorter than 4 days. Over the drip irrigated fields, the statistical metrics are R = 0.74, RMSE = 24.8 mm and bias = 2.3 mm for irrigation amounts cumulated over 15 days. When using SSM products derived from Sentinel-1 data, the statistical metrics on 15-day cumulated amounts slightly dropped to R = 0.64, RMSE = 28.7 mm and bias = 1.9 mm. The metrics on the seasonal amount retrievals are close to assimilating in situ observations with R = 0.99, RMSE = 33.5 mm and bias = −18.8 mm. Finally, among four rainfed seasons, only one false event was detected. This study opens perspectives for the regional retrieval of irrigation amounts and timing at the field scale and for mapping irrigated/non irrigated areas.

Making inference of British household's happiness efficiency: A Bayesian latent model

In this paper, we propose a novel approach whereby happiness for British households is identified within a latent model frontier analysis using longitudinal data. By doing so we overcome issues related to the measurement of happiness. To estimate happiness frontier and thereby happiness efficiency, we employ a Bayesian inference procedure organized around Sequential Monte Carlo (SMC) particle filtering techniques. In addition, we propose to consider individual-specific characteristics by estimating happiness efficiency models with individual-specific thresholds to happiness. This is the first study that treats happiness as a latent variable and departs from restrictions that happiness efficiency would be time invariant. Our results show that happiness efficiency is related to the welfare loss associated with potentially misusing the resources that British individuals have at their disposal. Key to happiness is to have certain personality traits, such as being agreeable and extravert as they assist efforts to enhance happiness efficiency. On the other hand, being neurotic impairs happiness efficiency.