The DOI matches its Crossref record; 35 of the 35 references it cites were found in a registry or an index.
Enhanced Deep Learning-Aided Kalman Filter Based on KalmanNet for Robust Nonlinear State Estimation
Abstract
Accurately estimating the real-time state of dynamic systems from noisy measurements is a fundamental task in signal processing. State-space models (SSMs) are commonly utilized to model system dynamics and account for uncertainty in both system state evolution and measurement data. Kalman filter (KF) and its variants are widely recognized for their low computational complexity and ability to address state evolution and measurement update. In practice, accurately modeling the system dynamics and obtaining model parameters are difficult, especially for nonlinear systems. These tasks often require predefined motion and measurement models that rely on coarse approximations and extensive manual tuning. In this article, we propose the deep-Learning-aided Kalman filter (DL-KF), a robust deep-learning-aided KF designed to perform KF in nonlinear dynamic environments where the process model is unavailable, and the system noise is also unknown. However, a linear measurement model is assumed. DL-KF utilizes a long short-term memory (LSTM) network to capture nonlinear process dynamics and generate state priors from historical measurements. Additionally, a gated recurrent unit (GRU) is utilized to independently learn innovation covariance, addressing measurement noise uncertainty, followed by the computation of the Kalman gain within the KF framework. By integrating an SSM framework with deep learning modules, the proposed method strikes the balance between the flexibility of a data-drive…
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Kalman-filterassisted deep regression models have been shown to reduce noise in RSSI sequences and achieve sub-meter indoor localization accuracy [9].”
From Child Playground Entry/Exit Tracking and Log Visualization System Using BLE Beacons and Smart Devices (2026) · cites this paper as [9]
Citations
Cites
Graves, A. (2012) · Studies in Computational Intelligence · doi:10.1007/978-3-642-24797-2_4 · Cited by 2,727
A Novel Adaptive Kalman Filter With Inaccurate Process and Measurement Noise Covariance Matrices
Huang, Y., Zhang, Y., Wu, Z., Li, N., & Chambers, J. (2018) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2017.2730480 · Cited by 731 · cited 1×
“When the distributions of process or measurement noise are unknown, approximate methods or adaptive filtering techniques are typically employed to estimate the noise statistics [8], [26].”
KalmanNet: Neural Network Aided Kalman Filtering for Partially Known Dynamics
Revach, G., Shlezinger, N., Ni, X., Escoriza, A. L., van Sloun, R. J. G., & Eldar, Y. C. (2022) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2022.3158588 · Cited by 599 · cited 10×
“The SSM is characterized by discrete-time, time-varying dynamics, similar to those described in [20] x ( k ) = F k ( x k − 1 ) x k − 1 + e k (17) where Here, X 1 , X 2 , and X 3 denote the horizontal temperature variation, the vertical temperature variation, and the convection rate of a fluid,…”
InfluentialUniversity of Michigan North Campus long-term vision and lidar dataset
Carlevaris-Bianco, N., Ushani, A. K., & Eustice, R. M. (2016) · The International Journal of Robotics Research · doi:10.1177/0278364915614638 · Cited by 576 · cited 2×
“Next, we extend the evaluation to the real-world NCLT dataset [32], which provides large-scale, multisensor data collected under diverse conditions.”
SegRNN: Segment Recurrent Neural Network for Long-Term Time-Series Forecasting
Lin, S., Lin, W., Wu, W., Zhao, F., Mo, R., & Zhang, H. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3647705 · Cited by 369 · cited 2×
“Recurrent neural networks (RNNs) have been used to capture temporal dependencies in nonlinear processes [16], while attention mechanisms and Transformer models provide a more flexible way to aggregate long-range context under complex observation conditions [17].”
A Short-Term Traffic Flow Prediction Model Based on an Improved Gate Recurrent Unit Neural Network
Shu, W., Cai, K., & Xiong, N. N. (2022) · IEEE Transactions on Intelligent Transportation Systems · doi:10.1109/tits.2021.3094659 · Cited by 262 · cited 1×
“We specifically chose the GRU network for processing these input features due to its computational e ffi ciency and fewer parameters compared to LSTM networks [30].”
DeepMTT: A deep learning maneuvering target-tracking algorithm based on bidirectional LSTM network
Liu, J., Wang, Z., & Xu, M. (2020) · Information Fusion · doi:10.1016/j.inffus.2019.06.012 · Cited by 181 · cited 1×
“Representative examples include DeepMTT, which learns maneuvering target motion from diverse observations for civil airport surveillance [18], and Transformer-based estimators explored for multitar-get tracking and filtering-style inference [19].”
Self-citationGao, B., Hu, G., Zhong, Y., & Zhu, X. (2021) · IEEE Sensors Journal · doi:10.1109/jsen.2021.3073963 · Cited by 147 · cited 1×
“Nonlinear variants, including the extended KF (EKF) [4], sigma-point filters such as the unscented KF (UKF) [5] and the cubature KF (CKF) [6], and particle filters (PFs) [7], extend KF to nonlinear settings.”
A Two-Step Parameter Optimization Method for Low-Order Model-Based State-of-Charge Estimation
Bian, X., Wei, Z., He, J., Yan, F., & Liu, L. (2020) · IEEE Transactions on Transportation Electrification · doi:10.1109/tte.2020.3032737 · Cited by 131 · cited 1×
“Nonlinear variants, including the extended KF (EKF) [4], sigma-point filters such as the unscented KF (UKF) [5] and the cubature KF (CKF) [6], and particle filters (PFs) [7], extend KF to nonlinear settings.”
A Slide Window Variational Adaptive Kalman Filter
Huang, Y., Zhu, F., Jia, G., & Zhang, Y. (2020) · IEEE Transactions on Circuits & Systems II Express Briefs · doi:10.1109/tcsii.2020.2995714 · Cited by 122 · cited 3×
“We also evaluate the proposed method against KF, least squares (LS), IMM, SWVAKF [33], AKNet [15], and KalmanNet on synthetic data generated by an SSM.”
Tightly Coupled Integration of GNSS, INS, and LiDAR for Vehicle Navigation in Urban Environments
Li, S., Wang, S., Zhou, Y., Shen, Z., & Li, X. (2022) · IEEE Internet of Things Journal · doi:10.1109/jiot.2022.3194544 · Cited by 112 · cited 2×
“A UTONOMOUS driving systems in global navigation satellite system (GNSS)-degraded environments such as long tunnels and severe occlusions require real-time vehicle tracking and localization using heterogeneous sensors (e.g., inertial, odometry, and radar / vision) [1], [2], [3].”
Split-KalmanNet: A Robust Model-Based Deep Learning Approach for State Estimation
Choi, G., Park, J., Shlezinger, N., Eldar, Y. C., & Lee, N. (2023) · IEEE Transactions on Vehicular Technology · doi:10.1109/tvt.2023.3270353 · Cited by 106 · cited 1×
“Following this idea, Split-KalmanNet [21] introduces a split-structure recurrent architecture.”
Latent-KalmanNet: Learned Kalman Filtering for Tracking From High-Dimensional Signals
Buchnik, I., Revach, G., Steger, D., van Sloun, R. J. G., Routtenberg, T., & Shlezinger, N. (2023) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2023.3344360 · Cited by 101 · cited 1×
“Latent-KalmanNet adds a representation module to compress high-dimensional observations into a compact latent space before filtering, reducing the burden of specifying complex measurement models [23].”
DynaNet: Neural Kalman Dynamical Model for Motion Estimation and Prediction
Chen, C., Lu, C. X., Wang, B., Trigoni, N., & Markham, A. (2021) · IEEE Transactions on Neural Networks and Learning Systems · doi:10.1109/tnnls.2021.3112460 · Cited by 85
Li, S., Li, Z., Li, J., Fernando, T. L., Iu, H. H., Wang, Q., & Liu, X. (2020) · IEEE Transactions on Industrial Electronics · doi:10.1109/tie.2020.2987279 · Cited by 77 · cited 1×
“Filtering-based solutions rely on an explicit state-space model (SSM) and require accurate noise statistics to compute filter gains [8], [9].”
Adaptive Kalmannet: Data-Driven Kalman Filter with Fast Adaptation
Ni, X.-Y., Revach, G., & Shlezinger, N. (2023) · IEEE International Conference on Acoustics, Speech, and Signal Processing · doi:10.1109/icassp48485.2024.10447012 · Cited by 59 · cited 5×
“We also evaluate the proposed method against KF, least squares (LS), IMM, SWVAKF [33], AKNet [15], and KalmanNet on synthetic data generated by an SSM.”
InfluentialAdaptive-Robust Fusion Strategy for Autonomous Navigation in GNSS-Challenged Environments
Shen, K., Li, Y., Liu, T., Zuo, J., & Yang, Z. (2023) · IEEE Internet of Things Journal · doi:10.1109/jiot.2023.3315758 · Cited by 51 · cited 2×
“Adaptive KF attempts to reduce this sensitivity by updating covariance or noise parameters online using innovations [3], [10].”
A New Coupled Method of SINS/DVL Integrated Navigation Based on Improved Dual Adaptive Factors
Liu, S., Zhang, T., Zhang, J., & Zhu, Y. (2021) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2021.3106118 · Cited by 48 · cited 1×
“To tackle the challenges in the state-space modeling, traditional methods, employed in [8], [27], and [28], include techniques such as worst-case error minimization and sys-tem identification, e.g., covariance matrix adaptation and expectation–maximization algorithms, to estimate the model…”
MAML-KalmanNet: A Neural Network-Assisted Kalman Filter Based on Model-Agnostic Meta-Learning
Chen, S., Zheng, Y., Lin, D., Cai, P., Xiao, Y., & Wang, S. (2025) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2025.3540018 · Cited by 46 · cited 1×
“Beyond gain-centric learning, MAML-KalmanNet incorporates model-agnostic meta-learning to enable fast adaptation across varying SSMs on limited task-specific data [22].”
Deep Learning for Model-Based Multiobject Tracking
Pinto, J., Hess, G., Ljungbergh, W., Xia, Y., Wymeersch, H., & Svensson, L. (2023) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2023.3289164 · Cited by 43 · cited 1×
“Representative examples include DeepMTT, which learns maneuvering target motion from diverse observations for civil airport surveillance [18], and Transformer-based estimators explored for multitar-get tracking and filtering-style inference [19].”
Next Generation Multitarget Trackers: Random Finite Set Methods vs Transformer-based Deep Learning
Pinto, J., Hess, G., Ljungbergh, W., Xia, Y., Svensson, L., & Wymeersch, H. (2021) · International Conference on Information Fusion (FUSION) · doi:10.23919/fusion49465.2021.9626990 · Cited by 36
EKFNet: Learning System Noise Covariance Parameters for Nonlinear Tracking
Xu, L., & Niu, R. (2024) · IEEE Transactions on Signal Processing · doi:10.1109/tsp.2024.3417350 · Cited by 35 · cited 1×
“To further evaluate the robustness of our method, we also compare DL-KF with other advanced neural filters, namely, DynaNet [35] and DeepKF (EKFNet) [36].”
Adaptive Kalman Filtering Based on Model Parameter Ratios
Ge, Q., Li, Y., Wang, Y., Xiaoming Hu, A., Li, H., & Sun, C. (2024) · IEEE Transactions on Automatic Control · doi:10.1109/tac.2024.3376306 · Cited by 34 · cited 2×
“Together with the linear-Gaussian measurement model, standard Bayesian filtering theory guarantees that the posterior p ( x k | z 1: k ) remains Gaussian and is uniquely characterized by the posterior mean and covariance given by the Kalman update equations [see (12)–(14)].”
Qiao, S., Fan, Y., Wang, G., Mu, D., & He, Z. (2022) · IEEE Transactions on Industrial Informatics · doi:10.1109/tii.2022.3220866 · Cited by 24 · cited 1×
“tational cost due to iterative posterior updates [13].”
Du, S., Huang, Y., Wen, W., & Zhang, Y. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3527555 · Cited by 19 · cited 2×
“It mixes multiple candidate models, improving robustness under maneuvers such as rapid acceleration and turning [14].”
Si, K., Li, P., Yuan, Z.-P., Qiao, K., Wang, B., & He, X. S. (2023) · IEEE Transactions on Instrumentation and Measurement · doi:10.1109/tim.2023.3293566 · Cited by 18 · cited 3×
“Filtering-based solutions rely on an explicit state-space model (SSM) and require accurate noise statistics to compute filter gains [8], [9].”
Shen, Z., Li, X., Zhou, Y., Li, S., Wu, Z., & Wang, X. (2024) · IEEE Transactions on Automation Science and Engineering · doi:10.1109/tase.2024.3422081 · Cited by 18 · cited 2×
“Parameter H rk is a known measurement matrix, and H rk = diag[1 , 1 , 1 , 1], v rk is an additive noise, which is a zero-mean white Gaussian noise with a known covariance matrix R rk .”
An Improved Sage-Husa Variational Robust Adaptive Kalman Filter With Uncertain Noise Covariances
Fan, Y., Qiao, S., Wang, G., & Zhang, H. (2024) · IEEE Sensors Journal · doi:10.1109/jsen.2024.3421271 · Cited by 18 · cited 1×
“The Sage–Husa (SH) adaptive KF is widely used, but the estimated process and measurement noise covariances may lose semipositive definiteness during operation [11].”
XGrad: Boosting Gradient-Based Optimizers With Weight Prediction
Guan, L., Li, D., Shi, Y., & Meng, J. (2024) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2024.3387399 · Cited by 15 · cited 1×
“To optimize the network parameters, we use the stochastic gradient descent algorithm [31].”
UKF-Based Optimal Tracking Control for Uncertain Dynamic Systems With Asymmetric Input Constraints
Liu, N., Zhang, K., Xie, X., & Yue, D. (2024) · IEEE Transactions on Cybernetics · doi:10.1109/tcyb.2024.3471987 · Cited by 14 · cited 2×
“Specifically, for SWVAKF, we tuned the sliding window size within the range of [5 , 50] and the forgetting factor ρ between 0.9 and 0.99 to balance tracking responsiveness against variational stability.”
InfluentialAn Adaptive Variational Outlier-Robust Filter for Multisensor Distributed Fusion
Fan, Y., Qiao, S., Wang, G., & Zhang, H. (2023) · IEEE Sensors Journal · doi:10.1109/jsen.2023.3339138 · Cited by 12 · cited 1×
“To tackle the challenges in the state-space modeling, traditional methods, employed in [8], [27], and [28], include techniques such as worst-case error minimization and sys-tem identification, e.g., covariance matrix adaptation and expectation–maximization algorithms, to estimate the model…”
Zhai, W., Cui, Y., Wang, L., Wang, N., Li, Z., Liu, P., & Hang, Z. (2025) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3597409 · Cited by 10 · cited 1×
“Recently, attention-based architectures have also been coupled with Kalman-style recursion, for example, by using Transformers to learn Kalman gain updates or by analyzing structural connections between Transformer inference and Kalman filtering [24], [25].”
An Adaptive Extended Kalman Filter Based on Variational Bayesian and Sage-Husa Prediction Algorithms
Ma, L., Cao, R., & Huang, C. L. (2024) · International Conference on Information Communication and Signal Processing (ICICSP) · doi:10.1109/icicsp62589.2024.10809212 · Cited by 6 · cited 1×
“Adaptive KF attempts to reduce this sensitivity by updating covariance or noise parameters online using innovations [3], [10].”
Xin, X., Zhou, H., Feng, X., Pan, F., Wang, J., & Li, Z. (2026) · IEEE Internet of Things Journal · doi:10.1109/jiot.2025.3650118 · Cited by 2 · cited 1×
“Recently, attention-based architectures have also been coupled with Kalman-style recursion, for example, by using Transformers to learn Kalman gain updates or by analyzing structural connections between Transformer inference and Kalman filtering [24], [25].”
A Fast Particle Filter for Tracking an Uncooperative GNSS Interference Source
Biswas, S. K., & Cetin, E. (2025) · IEEE Transactions on Aerospace and Electronic Systems · doi:10.1109/taes.2025.3579487 · Cited by 1 · cited 1×
“Nonlinear variants, including the extended KF (EKF) [4], sigma-point filters such as the unscented KF (UKF) [5] and the cubature KF (CKF) [6], and particle filters (PFs) [7], extend KF to nonlinear settings.”
Research works printed but not linked: 1
- Bar-Shalom, 2011, Tracking and Data Fusion
Cited by
Gargouri, S., Mansouri, M. M., Kahloul, A. A., Kermani, M., & Sakly, A. (2026) · Energies · doi:10.3390/en19143301 · Cited by 1
Qian, B., Huang, L., & Guo, X. (2026) · International Conference on Intelligent Computing and Signal Processing (ICSP) · doi:10.1109/icsp69961.2026.11540802
“Existing studies have enhanced filtering processes from perspectives such as sample screening [3], deep learning-assisted modeling [4], and adaptive estimation of unknown parameters [5], thereby improving adaptability to complex environments to some extent.”
GMT: A Graph-Based State Estimation Framework via Improved Graph Neural Networks
Song, Q., & Yang, Z. (2026) · International Conference on Image Processing, Multimedia Technology and Machine Learning (IPMML) · doi:10.1109/ipmml71439.2026.11682445
“Traditional state estimation methods rely on accurate physical equations or mechanistic models, achieving state evolution through explicit mathematical formulations [5]–[7].”
Gao, R. (2026) · Applied Sciences · doi:10.3390/app162110444
Metadata
- Authors
- Fu, T., Mao, G., Wang, Z., & Wang, K.
- Affiliations
- Xidian University (CN)
- Southeast University (CN)
- Venue
- IEEE Internet of Things Journal
- Published
- 2026-06-01
- Volume
- 13
- Issue
- 11
- Pages
- 24369-24386
- ISSN
- 2327-4662 (ISSN-L)
- 2372-2541
- OpenAlex
- W7134858622
- Cited by
- 6
- References
- 35
- Citations by year
- 2026: 6
- FWCI
- 20.50
- Type
- journal article
- Language
- en
- Open access
- closed
- Licence
- Locations
- IEEE Internet of Things Journal (publishedVersion)
- Funders
- National Natural Science Foundation of China (U21A20446)
- Wuxi Key Laboratory of Integrated Vehicle-Road-Cloud Smart Transportation System
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Target Tracking and Data Fusion in Sensor Networks
- Physical Sciences › Engineering › Control and Systems Engineering › Control Systems and Identification
- Physical Sciences › Engineering › Aerospace Engineering › Inertial Sensor and Navigation
- Keywords
- KalmanNet, nonlinear state estimation, state-space model, Kalman gain, unknown dynamics, Lorenz attractor
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.