Prediction Filter Research Articles

Robustness Stability Analysis of Higher-Order DPCM Prediction Filters

This paper considers the robustness of the differential pulse-code modulation system with higher-order predictors. Special attention is paid to the robust parametric stability of the prediction filters with respect to the predictor coefficients. A generalisation of robustness in the classical sense is performed, and appropriate relations for calculating the probability of robustness are derived using Kharitonov principle. The proposed robustness estimation method is used for the third- and fourth-order prediction filters on speech signals, where the application of traditional methods is too difficult. For this reason, the Monte Carlo method is used to solve complex probability integrals. Verification and error analysis are performed for the previously considered second-order predictor. Satisfactory predetermined accuracy is achieved by increasing the number of samples. The results obtained could be very useful to design a system with suitable values for the predictor coefficients.

Modular control architecture for safe marine navigation: Reinforcement learning with predictive safety filters

Many autonomous systems are safety-critical, making it essential to have a closed-loop control system that satisfies constraints arising from underlying physical limitations and safety aspects in a robust manner. However, this is often challenging to achieve for real-world systems. For example, autonomous ships at sea have nonlinear and uncertain dynamics and are subject to numerous time-varying environmental disturbances such as waves, currents, and wind. There is increasing interest in using machine learning-based approaches to adapt these systems to more complex scenarios, but there are few standard frameworks that guarantee the safety and stability of such systems. Recently, predictive safety filters (PSF) have emerged as a promising method to ensure constraint satisfaction in learning-based control, bypassing the need for explicit constraint handling in the learning algorithms themselves. The safety filter approach leads to a modular separation of the problem, allowing the use of arbitrary control policies in a task-agnostic way. The filter takes in a potentially unsafe control action from the main controller and solves an optimization problem to compute a minimal perturbation of the proposed action that adheres to both physical and safety constraints. In this work, we combine reinforcement learning (RL) with predictive safety filtering in the context of marine navigation and control. The RL agent is trained on path-following and safety adherence across a wide range of randomly generated environments, while the predictive safety filter continuously monitors the agents' proposed control actions and modifies them if necessary. The combined PSF/RL scheme is implemented on a simulated model of Cybership II, a miniature replica of a typical supply ship. Safety performance and learning rate are evaluated and compared with those of a standard, non-PSF, RL agent. It is demonstrated that the predictive safety filter is able to keep the vessel safe, while not prohibiting the learning rate and performance of the RL agent.

Online bone/air-conducted speech fusion in the presence of strong narrowband noise

Online adaptive bone/air-conducted speech fusion is a very promising and attractive technique for real time speech communication in noisy environments. This paper firstly demonstrates that conventional bone/air-conducted speech fusion systems may degrade significantly, despite that only a tiny narrowband noise pollution persists in bone-conducted speech by extensive simulations and experiments. To address this issue, a robust nonlinear system is proposed that successfully enhances the speech contaminated by the harsh narrowband noise. The penetrated narrowband noise is extracted by an adaptive linear prediction filter from an extra reference signal that is capture by a non-contacted bone sensor, and then fed to the parallel adaptive noise cancellers to reconstruct a cleaner error feedback for the speech recovery system. Due to this combination, the proposed system enjoys both attractive robustness against the harsh additive noise and strong real-time capability. Simulations as well as application to a real factory noise are conducted to demonstrate the superior performance of the proposed system.

A hybrid information fusion method for SINS/GNSS integrated navigation system utilizing GRU-aided AKF during GNSS outages

Abstract To enhance the performance of integrated inertial navigation system (INS) and global navigation satellite system (GNSS) during GNSS outages, this paper proposed a fusion positioning method based on predictive observation information and adaptive filter parameter. Combined with an adaptive Kalman filter (AKF) and a Gated Recurrent Unit neural network (NN) that directly relates the inertial measurement unit (IMU) output sequence to the error estimation, the hybrid information fusion system can provide effective corrections to compensate for horizontal position errors under the constraints of complex and dynamic vehicle movement data during GNSS outages. Meanwhile, the designed adaptive parameter of the integrated navigation filter can adjust the credibility of the state prediction section when the GNSS is reconnected, ensuring the system can switch rapidly between the INS/GNSS and INS/NN integrated modes. The performance of the proposed information fusion method has been experimentally validated using IMU and GNSS data collected in a vehicle navigation test conducted on a stretch of expressway. The comparison results indicate that the proposed algorithm has error suppression capabilities under various experimental constraints and demonstrates a degree of extendibility and reusability.

Intelligent filter to remove artifact during cryo freezing for pulmonary vein isolation increases the time window for observation of time of isolation

Abstract Introduction Time to pulmonary vein isolation (PVI) during cryo balloon (CB) PVI is a critical data point in characterizing the pulmonary vein (PV) isolation status as it improves clinical outcomes. PV potentials on the circular mapping catheter may be obscured by power line artifact during the freeze process as ice forms near the electrode. This study assesses the ability of an intelligent digital filter to sample, predict and remove noise from PV electrograms during the freeze cycle to increase the time window available to observe the abolition of PV potentials. Method An intelligent filter that detects and removes power line related signal artifact was developed. The filter algorithm was then retrospectively applied to electrograms collected during delivery of cryo energy in the PVISION trial of the PVI Analyzer study previously described. The PV potential amplitude immediately before the delivery of cryo energy was recorded. During the freeze cycle, artifact amplitude was measured and the time required for the noise to exceed that of the original PV potential recorded. This operation was repeated with the application of the predictive filter. An example of an analysis of a single bipolar trace is shown in Fig 1. Results The PVISION trial included 47 patients who underwent a cryo PVI procedure (42% female, mean age 59 years, +/- 11 years). Of those, 24 cases and 68 vein cryo applications had sufficient information about when the cryo freeze cycle started and ended. Of these 68 veins treated, 33 exhibited power line noise on one or more bipolar traces during the freeze cycle (49%). The mean amplitude of PV potentials before application of energy was 0.49 mV (+/- 0.79 mV) and the mean time to artifact exceeding this level was 69 (+/-50 seconds), across all bipolar traces with freeze noise present. After application of the intelligent filter, the mean time to artifact exceeding the baseline PV amplitude increased to 82 seconds (+/- 57 seconds), across all bipolar traces with freeze noise present. In 44 veins noise amplitude never exceeded PV potential amplitude on any of the bipolar traces during the freeze after application of the intelligent filter. Paired analysis of the average time of noise exceeding the PV potential amplitude during the freeze was 116 s before application of the filter and 26 s with the filter applied. See Fig 2. for a summary of the results. Conclusion An intelligent digital filter designed to remove artifact from electrograms during cryo ablation procedures might increase the time window before noise saturates PV potentials and, in some cases, remove noise such that it never exceeds that of the PV potential.Fig 1 Example of noise and filteringFig 2. Summary of results

Confidence Weighted Learning Entropy for Fault-Tolerant Control of a PMSM With a High-Resolution Hall Encoder

There are several non-idealities that can degrade magnetic Hall-effect sensors performance and impact related applications. Thus, a confidence weighted learning entropy (CWLE) is proposed as a fault-tolerant control strategy for field-oriented control (FOC) of permanent magnet synchronous machines (PMSM). It combines sensorless and sensor-based control, while capitalizing on their major advantages, such as operation from standstill and at lower speeds, fast dynamic response, and fault tolerance to encoder errors. Encoder fault detection is based on learning entropy, that monitors weights increments of two predictive filters of angular displacement. If the two observed systems behave similarly, the variance of the weights increments is similar. A higher variance, on the other hand, reflects unforeseen misbehavior of a particular system, leading to a decrease in its confidence. A voting mechanism based on confidence weighted average then decides which of the two systems should be used for FOC of PMSM. The method has been experimentally verified on a high-speed PMSM, achieving more reliable control performance with fast response to encoder failures.

Active control method for fluid borne noise in aerospace fluid systems of variable operation statuses

In the fluid system of aerospace vehicles, the fluid-borne noise (FBN) may induce vibration and damage to components. The FBN may limit the performance of the entire onboard system. To achieve a stable active elimination for the FBN is not easy, for many variable operation statuses exist in practical systems, such as varying-frequency and varying-load status. In this work, an improved feed-forward active control method based on narrowband adaptive notch filters, with an online identification module, is proposed. The algorithm includes three adaptive filters: an adaptive notch filter for producing cancelling signal by using filtered-X least mean square (FXLMS) algorithm, one for online identification to model the real-time secondary path, and an error prediction filter for extracting the precise error for online identification. In the aircraft hydraulic system of high pressure, a so-called relief scheme, also known as the bypass structure, is used for the FNB, controlling an active control valve in the branched line to overflow the main fluid line. The system is modelled with software AMESim for the fluid part and Matlab for the control algorithm. Simulation results show that the control method can performance well in the situations of varying frequency and load. The test rig of a hydraulic system is established, which could change pump speed and fluid pressure. The test results show that more than 90% reduction for a certain frequency can be achieved for stable pressure statuses. And more than 50% reduction is achieved for varying frequency with step changes. As for the load-varying status, the reduction of 82.7% and 65.8% are obtained before and after pressure change. The method proposed is validated to be effective with both numerical and experimental tests.

Benchmarking machine learning-based real-time respiratory signal predictors in 4D SBRT.

Stereotactic body radiotherapy of thoracic and abdominal tumors has to account for respiratory intrafractional tumor motion. Commonly, an external breathing signal is continuously acquired that serves as a surrogate of the tumor motion and forms the basis of strategies like breathing-guided imaging and gated dose delivery. However, due to inherent system latencies, there exists a temporal lag between the acquired respiratory signal and the system response. Respiratory signal prediction models aim to compensate for the time delays and to improve imaging and dosedelivery. The present study explores and compares six state-of-the-art machine and deep learning-based prediction models, focusing on real-time and real-world applicability. All models and data are provided as open source and data to ensure reproducibility of the results and fosterreuse. The study was based on 2502 breathing signals ( h) acquired during clinical routine, split into independent training (50%), validation (20%), and test sets (30%). Input signal values were sampled from noisy signals, and the target signal values were selected from corresponding denoised signals. A standard linear prediction model (Linear), two state-of-the-art models in general univariate signal prediction (Dlinear, Xgboost), and three deep learning models (Lstm, Trans-Enc, Trans-TSF) were chosen. The prediction performance was evaluated for three different prediction horizons (480, 680, and 920ms). Moreover, the robustness of the different models when applied to atypical, that is, out-of-distribution (OOD) signals, was analyzed. The Lstm model achieved the lowest normalized root mean square error for all prediction horizons. The prediction errors only slightly increased for longer horizons. However, a substantial spread of the error values across the test signals was observed. Compared to typical, that is, in-distribution test signals, the prediction accuracy of all models decreased when applied to OOD signals. The more complex deep learning models Lstm and Trans-Enc showed the least performance loss, while the performance of simpler models like Linear dropped the most. Except for Trans-Enc, inference times for the different models allowed for real-time application. The application of the Lstm model achieved the lowest prediction errors. Simpler prediction filters suffer from limited signal history access, resulting in a drop in performance for OODsignals.

Random noise attenuation of 3D multicomponent seismic data using a fast adaptive prediction filter

Random noise in seismic records affects the accuracy of effective signal identification, making it difficult for subsequent seismic data processing, imaging, and interpretation. Therefore, random noise attenuation has always been an important step in seismic data processing, especially for 3D data. In recent years, multicomponent exploration has been developed rapidly. However, the common method for processing multicomponent data is to process each component separately resulting in the correlation between multicomponent data being neglected. For 3D multicomponent data, we develop a multicomponent adaptive prediction filter (MAPF) based on noncausal regularized nonstationary autoregressive models to implement random noise attenuation in the t- x- y domain. The MAPF for multicomponent signals can be used to identify the potential correlations and differences between each pair of components, providing not only a robust analysis of the individual components but also effective information about the consistency and differences between each component with more information and constraints compared with traditional 1C prediction. Moreover, it can obtain smooth nonstationary prediction coefficients by solving the least-squares problem with shaping regularization. The example results demonstrate that the MAPF method is superior to the traditional adaptive prediction filtering method. Furthermore, because the multicomponent method requires more coefficients and takes a longer time to predict than the 1C method, we further develop a fast MAPF (FMAPF) combining the data pooling and coefficient reconstruction strategies. The example results demonstrate that the FMAPF method is effective at denoising and greatly improves computational efficiency. The method comes with a slight decrease in computational accuracy.

Predictive stability filters for nonlinear dynamical systems affected by disturbances

Predictive safety filters provide a way of projecting potentially unsafe inputs, proposed, e.g. by a human or learning-based controller, onto the set of inputs that guarantee recursive state and input constraint satisfaction by leveraging model predictive control techniques. In this paper, we extend this framework such that in addition, robust asymptotic stability of the closed-loop system can be guaranteed by enforcing a decrease of an implicit Lyapunov function which is constructed using a predicted system trajectory. Differently from previous results, we show robust asymptotic stability with respect to a predefined disturbance set on an extended state consisting of the system state and a warmstart input sequence. The proposed strategy is applied to an automotive lane keeping example in simulation.

Learning safe control for multi-robot systems: Methods, verification, and open challenges

In this survey, we review the recent advances in control design methods for robotic multi-agent systems (MAS), focusing on learning-based methods with safety considerations. We start by reviewing various notions of safety and liveness properties, and modeling frameworks used for problem formulation of MAS. Then we provide a comprehensive review of learning-based methods for safe control design for multi-robot systems. We start with various shielding-based methods, such as safety certificates, predictive filters, and reachability tools. Then, we review the current state of control barrier certificate learning in both a centralized and distributed manner, followed by a comprehensive review of multi-agent reinforcement learning with a particular focus on safety. Next, we discuss the state-of-the-art verification tools for the correctness of learning-based methods. Based on the capabilities and the limitations of the state-of-the-art methods in learning and verification for MAS, we identify various broad themes for open challenges: how to design methods that can achieve good performance along with safety guarantees; how to decompose single-agent-based centralized methods for MAS; how to account for communication-related practical issues; and how to assess transfer of theoretical guarantees to practice.

Thermally efficient gas turbine with pressure drop-based automated filter cleaning and optimized fuel control system

Optimizing elements like fuel specifications and inlet–outlet conditions is essential for enhancing gas turbine performance. Prior research optimized filter house pressure drop using statistical methods, but these methods had drawbacks related to measurement errors. Additionally, with changing air conditions, periodic-based cleaning for self-cleaning pulse filters at the turbine inlet proved unsuccessful, making the procedure difficult and time-consuming. A novel “Thermally Efficient Gas Turbine with Pressure Drop-Based Automated Filter Cleaning and Optimized Fuel Control System” has been suggested to improve gas turbine performance. To handle pressure drop effectively, this cutting-edge system adds a predictive Apriori-based parallel auto-cleaning pulse filter, with automatic cleaning activated at a predetermined pressure difference. An adaptive neuro-fuzzy inference system-multi-layer perceptron integrated simulated annealing-Levenberg Marquardt algorithm has been created to handle mistakes in prediction and improve turbine parameters for greater energy efficiency. This complete strategy considerably increases energy efficiency while also improving air pressure drop prediction. A 200 MW power output, 70 % energy efficiency, 1.3 kg/kW-h fuel consumption, a small residual error (−0.4 mbar), and a pressure difference of only 3 mbar are just a few of the model's excellent performance indicators.

Road slope estimation for heavy-duty vehicles under the influence of multiple source factors in real complex road environments

Accurate road slope information is of paramount importance for optimizing vehicle control and enhancing energy efficiency. In the context of heavy-duty vehicles (HDVs), the precise estimation of road slope presents a formidable challenge. This challenge is rooted in the coupling between vehicle mass and road slope, the robustness of road slope estimation methods across diverse driving styles, and the generalizability of algorithms to real-world road scenarios. To address these challenges, this paper proposes an innovative road slope estimation scheme that comprehensively considers the influences of multiple source factors in a real-world HDV transportation scenario. Firstly, in order to solve the problem of road slope estimation error under complex road types and different driving styles, a complex road slope estimation method based on the nonlinear predictive filter (NPF) combined with the Cubature Kalman filter (CKF) is constructed. To eliminate the problem of vehicle power fluctuation in special traffic scenarios, we propose a special scenario road slope estimation method based on the combination of a low-pass filter (LPF) and the NPF_CKF algorithm. To address the error problem when using low-cost Inertial Measurement Unit (IMU) sensors, we designed a Kalman filter (KF)-based multi-sensor fusion correction method. Finally, we conduct real-world experiments on complex roads at the test site to verify the accuracy of the proposed estimation method and its insensitivity to model parameters. The robustness and applicability of the proposed method are also verified in free-driving experiments conducted in real-world road scenarios.

Frequency estimation of multiple complex sinusoids using noise suppressing predictive FIR filter

The paper presents a predictive filter based frequency estimation algorithm (PF-FE) for multiple complex sinusoids corrupted by additive white Gaussian noise. The proposed frequency estimator combines the noise suppression capability of an FIR filter with a least squares approach minimizing the sum of the squared errors of a prediction filter. The FIR filter coefficients are chosen such that the sinusoidal components are predicted with minimum distortion. Computer simulation results are given to contrast the performance of the proposed PF-FE method with six other state-of-the-art frequency-estimators and the Cramer-Rao lower bound. Furthermore, the paper provides a complexity analysis of the PF-FE method and compares it with those of the conventional MUltiple SIgnal Classification algorithm and some selected state-of-the-art methods from the literature. Performance of the proposed PF-FE method is validated with effective mean square error (EMSE) plots for model orders of three and five and under different observation lengths and signal-to-noise ratio values. Finally, the paper demonstrates how EMSE varies for the proposed method when the prediction filter length is increased from 20 to 75. After extensive simulations, it is observed that when filter length is larger than half the observation length the EMSE will start to degrade.

Model-less prediction filter for adaptive adjustment process noise.

In this study, a filtering scheme suitable for high-precision sensors was proposed to extract high-precision sensor information. According to the principle of Kalman gain based on data fusion, a model-less prediction filter with minimum gain measurement noise compensation and process noise posteriori constraint adjustment was developed. In comparison to various Kalman filter methods, the proposed algorithm demonstrated better accuracy in the steady state. The high precision performance and effectiveness of the model-less prediction filter were verified under a digitally controlled linear power supply.

Reconstruction of Compressed Hyperspectral Image Using SqueezeNet Coupled Dense Attentional Net

This study addresses image denoising alongside the compression and reconstruction of hyperspectral images (HSIs) using deep learning techniques, since the research community is striving to produce effective results to utilize hyperspectral data. Here, the SqueezeNet architecture is trained with a Gaussian noise model to predict and discriminate noisy pixels of HSI to obtain a clean image as output. The denoised image is further processed by the tunable spectral filter (TSF), which is a dual-level prediction filter to produce a compressed image. Subsequently, the compressed image is analyzed through a dense attentional net (DAN) model for reconstruction by reverse dual-level prediction operation. All the proposed mechanisms are employed in Python and evaluated using a Ben-Gurion University-Interdisciplinary Computational Vision Laboratory (BGU-ICVL) dataset. The results of SqueezeNet architecture applied to the dataset produced the denoised output with a Peak Signal to Noise Ratio (PSNR) value of 45.43 dB. The TSF implemented to the denoised images provided compression with a Mean Square Error (MSE) value of 8.334. Subsequently, the DAN model executed and produced reconstructed images with a Structural Similarity Index Measure (SSIM) value of 0.9964 dB. The study proved that each stage of the proposed approach resulted in a quality output, and the developed model is more effective to further utilize the HSI. This model can be well utilized using HSI data for mineral exploration.

Predictive Control Barrier Functions: Enhanced Safety Mechanisms for Learning-Based Control

While learning-based control techniques often outperform classical controller designs, safety requirements limit the acceptance of such methods in many applications. Recent developments address this issue through so-called predictive safety filters, which assess if a proposed learning-based control input can lead to constraint violations and modifies it if necessary to ensure safety for all future time steps. The theoretical guarantees of such predictive safety filters rely on the model assumptions and minor deviations can lead to failure of the filter putting the system at risk. This paper introduces an auxiliary soft-constrained predictive control problem that is always feasible at each time step and asymptotically stabilizes the feasible set of the original safety filter, thereby providing a recovery mechanism in safety-critical situations. This is achieved by a simple constraint tightening in combination with a terminal control barrier function. By extending discrete-time control barrier function theory, we establish that the proposed auxiliary problem provides a ‘predictive’ control barrier function. The resulting algorithm is demonstrated using numerical examples.

Online Learning of the Kalman Filter With Logarithmic Regret

In this paper, we consider the problem of predicting observations generated online by an unknown, partially observable linear system, which is driven by Gaussian noise. In the linear Gaussian setting, the optimal predictor in the mean square error sense is the celebrated Kalman filter, which can be explicitly computed when the system model is known. When the system model is unknown, we have to learn how to predict observations online based on finite data, suffering possibly a non-zero regret with respect to the Kalman filter's prediction. We show that it is possible to achieve a regret of the order of <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\text{poly}\log (\mathsf {N})$</tex-math></inline-formula> with high probability, where <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$N$</tex-math></inline-formula> is the number of observations collected. This is achieved using an online least-squares algorithm, which exploits the approximately linear relation between future observations and past observations. The regret analysis is based on the stability properties of the Kalman filter, recent statistical tools for finite sample analysis of system identification, and classical results for the analysis of least-squares algorithms for time series. Our regret analysis can also be applied to other predictors, e.g. multiple step-ahead prediction, or prediction under exogenous inputs including closed-loop prediction. A fundamental technical contribution is that our bounds hold even for the class of non-explosive systems (including marginally stable systems), which was not addressed before in the case of online prediction.

Adaptive feedback cancellation based on interaural level difference using lattice filter with correlation control for binaural hearing devices

This paper proposes a correlation control method for linear prediction filters based on howling detection using binaural information in hearing aids. In a closed-loop system, high autocorrelation of the input signal can cause estimation errors owing to biased solutions, resulting in an acoustic artifact known as entrainment. Adaptive feedback cancellation (AFC) based on the linear prediction error method (PEM) has been successfully used as a solution to this bias. However, if the convergence speed of the linear prediction filter is faster than the estimated speed of the feedback path, removal of the feedback signal would be difficult. The proposed method uses a strategy to improve convergence by detecting howling based on the interaural level difference (ILD) and controlling the decorrelation performance of the correlation control algorithm of the adaptive lattice filter. Experimental results indicate that the adaptive filter of the proposed method significantly improves the estimation accuracy and tracking performance. The robustness of the proposed method is also confirmed via objective and subjective evaluation experiments on the generation of artifacts. Experimental results obtained using autocorrelated input signals show that the proposed method significantly improves robustness to entrainment compared to the conventional bilateral AFC method.

Multichannel Sparse Deconvolution of Teleseismic Receiver Functions With f − x Preconditioning

AbstractTeleseismic receiver functions (RFs) are an effective tool for structural imaging due to their simplicity and sensitivity to changes in rock elastic properties. Generally, RFs are estimated through trace‐by‐trace deconvolution. This individualized approach ignores the lateral coherency among RFs from neighboring raypaths, which could lead to rougher or incorrect seismic sections due to the presence of noise. This study presents a multichannel sparse deconvolution method that takes advantage of the cross‐trace coherency of RFs at individual stations for stable and accurate imaging outcomes. The proposed algorithm incorporates sparse inversion and frequency‐space prediction filters, which facilitate the retrieval of high‐resolution and spatially continuous conversion energy. Rigorous testing using synthetic and actual data typically suggests the higher robustness of multichannel deconvolution over single‐channel deconvolution, especially under low signal‐to‐noise ratios. The noise‐resistance property of the multichannel approach offers new opportunities to increase the volume of usable, high‐quality data for receiver function analysis.