Mihkel Tommingas
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Mihkel Tommingas has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Ivo Müürsepp
Thomas Johann Seebeck Department of Electronics, Tallinn University of Technology, Tallinn, Estonia
3 shared publications - Muhammad Mahtab Alam
Tallinn University of Technology
3 shared publications - Sander Varbla
Thomas Johann Seebeck Department of Electronics, Tallinn University of Technology, Tallinn, Estonia
2 shared publications - Aleksander Klauson
Tallinn University of Technology
1 shared publication - Julia Berdnikova
Thomas Johann Seebeck Dept. of Electronics, Tallinn Univ. of Technol., Ehitajate tee 5, Tallinn 19086, Estonia
1 shared publication - Mirko Mustonen
Tallinn University of Technology
1 shared publication
Latest Publications
- Open Access
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 6, 4383
Erratum to “UWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation”
Presents corrections to the paper, (Erratum to “UWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation”).
- 2 June 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 6, 4383
Open AccessErratum to “UWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation”
Presents corrections to the paper, (Erratum to “UWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation”).
- Open Access
- 26 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 6, 2177-2189
UWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation
This article presents a novel machine learning (ML) augmented sensor fusion scheme for seamless indoor-outdoor positioning using ultra-wideband (UWB) and global navigation satellite system (GNSS) sensors. Dilution of precision (DOP) is a common metric that describes the level of geometrical uncertainty and is often applied in sensor fusion schemes. However, it does not fully reflect other factors that may influence positioning integrity, such as signal quality, pseudorange error, or the number of servicing nodes. An incorrect estimation of sensor uncertainty may significantly affect the precision and robustness of individual sensors and their fused positioning solution, especially in indoor-outdoor transition zones, where positioning is most challenging. Therefore, this article proposes a sensor fusion scheme augmented with two distinct extreme gradient boosted (XGBoost) ML models for estimating UWB and GNSS positioning uncertainties. Trained using real-life datasets, these models have the advantage of considering an ensemble of features rather than a single parameter to estimate the uncertainty of the current coordinate. In contrast to sensor fusion solutions, which implement only highly accurate RTK fixed solution mode, the GNSS model can operate with different correction qualities as well. The proposed scheme was applied on a moving testbed with UWB and GNSS sensors while relying on the ML models to enhance the coordinate filtering. The results show that the ML-based approach can improve seamless transition between indoor and outdoor areas with almost no sensor dropouts with a mean positioning error of 0.16 m and a maximum error of approximately 0.5 m.
- 26 March 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Open Journal of the Communications Society
- Vol. 6, 2177-2189
Open AccessUWB and GNSS Sensor Fusion Using ML-Based Positioning Uncertainty Estimation
This article presents a novel machine learning (ML) augmented sensor fusion scheme for seamless indoor-outdoor positioning using ultra-wideband (UWB) and global navigation satellite system (GNSS) sensors. Dilution of precision (DOP) is a common metric that describes the level of geometrical uncertainty and is often applied in sensor fusion schemes. However, it does not fully reflect other factors that may influence positioning integrity, such as signal quality, pseudorange error, or the number of servicing nodes. An incorrect estimation of sensor uncertainty may significantly affect the precision and robustness of individual sensors and their fused positioning solution, especially in indoor-outdoor transition zones, where positioning is most challenging. Therefore, this article proposes a sensor fusion scheme augmented with two distinct extreme gradient boosted (XGBoost) ML models for estimating UWB and GNSS positioning uncertainties. Trained using real-life datasets, these models have the advantage of considering an ensemble of features rather than a single parameter to estimate the uncertainty of the current coordinate. In contrast to sensor fusion solutions, which implement only highly accurate RTK fixed solution mode, the GNSS model can operate with different correction qualities as well. The proposed scheme was applied on a moving testbed with UWB and GNSS sensors while relying on the ML models to enhance the coordinate filtering. The results show that the ML-based approach can improve seamless transition between indoor and outdoor areas with almost no sensor dropouts with a mean positioning error of 0.16 m and a maximum error of approximately 0.5 m.
- Open Access
- 24 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Indoor and Seamless Positioning and Navigation
- Vol. 2, 205-218
UWB Positioning Integrity Estimation Using Ranging Residuals and ML Augmented Filtering
This article investigates the use of Ultra-Wideband (UWB) ranging residuals for coordinate integrity estimation and their use in a filtering scheme. Typically, UWB system accuracy is improved using channel statistics to detect and mitigate non-line of sight (NLOS) effects between UWB sensors and the object to be located, potentially improving the end coordinate solution. However, in practice when considering UWB system with a high positioning update rate, this is not a feasible approach, as gathering and processing CS data takes too much time. In contrast to this approach, this article proposes a set of features based on UWB ranging residuals that could be used as an alternative in integrity assessment. By using machine learning (ML), the most important features were extracted from the initial set and then used to train and validate a model for UWB coordinate error prediction. Lastly, the prediction was applied in an adaptive Kalman filtering (AKF) scheme as an input for measurement uncertainty. Model testing was done using UWB measurement test dataset gathered at an industrial site. The overall results showed significant improvement in 2D and 3D positioning metrics of ML-augmented filtering when compared to non-ML filtering. On average, the end coordinates in the test set had ca. 10 cm smaller mean location error and ca. 40 cm smaller dispersion in 2D positioning. Additionally, the presence of outliers was reduced significantly as the maximum error offset decreased by several meters. Although ML augmented filtering is computationally slower than non-ML filtering (e.g., ordinary and extended Kalman filter), it is still faster than using channel statistics for UWB integrity estimation. The results show that using the proposed residual features in an ML model provides a feasible approach to predict UWB positioning integrity and use it as a measure of uncertainty in a coordinate filtering scheme.
- 24 June 2024
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Journal of Indoor and Seamless Positioning and Navigation
- Vol. 2, 205-218
Open AccessUWB Positioning Integrity Estimation Using Ranging Residuals and ML Augmented Filtering
This article investigates the use of Ultra-Wideband (UWB) ranging residuals for coordinate integrity estimation and their use in a filtering scheme. Typically, UWB system accuracy is improved using channel statistics to detect and mitigate non-line of sight (NLOS) effects between UWB sensors and the object to be located, potentially improving the end coordinate solution. However, in practice when considering UWB system with a high positioning update rate, this is not a feasible approach, as gathering and processing CS data takes too much time. In contrast to this approach, this article proposes a set of features based on UWB ranging residuals that could be used as an alternative in integrity assessment. By using machine learning (ML), the most important features were extracted from the initial set and then used to train and validate a model for UWB coordinate error prediction. Lastly, the prediction was applied in an adaptive Kalman filtering (AKF) scheme as an input for measurement uncertainty. Model testing was done using UWB measurement test dataset gathered at an industrial site. The overall results showed significant improvement in 2D and 3D positioning metrics of ML-augmented filtering when compared to non-ML filtering. On average, the end coordinates in the test set had ca. 10 cm smaller mean location error and ca. 40 cm smaller dispersion in 2D positioning. Additionally, the presence of outliers was reduced significantly as the maximum error offset decreased by several meters. Although ML augmented filtering is computationally slower than non-ML filtering (e.g., ordinary and extended Kalman filter), it is still faster than using channel statistics for UWB integrity estimation. The results show that using the proposed residual features in an ML model provides a feasible approach to predict UWB positioning integrity and use it as a measure of uncertainty in a coordinate filtering scheme.
- 1 October 2017
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 142 (4_Suppleme) , 2686
Underwater ship noise pattern detection and identification
The anthropogenic underwater noise event could be identified by spectral analysis and acoustic pattern pre-classification of previously measured noise sources. This paper discusses real-time applicable methods for separation of purely natural noise recordings when ships are absent from the data polluted by the ship noise. The natural noise periods are used for the statistical modeling of underwater channel environment and relative levels of marine ambient noise (Wenz curves). The anthropogenic noise periods allow us to identify the noise sources. Moreover, most of the ship identification methods require spectrum component processing and pre-determined ship classification, which could be based on this automatic separation results. The correlational and multiple criteria detection methods compared with cyclostationary feature detection and widely used energy detection. False alarm and misclassification minimization could be achieved by multiple sensor data fusion: weather and Automatic Identification System (AIS). The measured data were collected near the seafloor with an autonomous long-term passive acoustic recorder and were combined with ship passage information from the AIS. Three years of measurements include multiple individual ship observations with different speeds and directions. Analysis of the experimental results emphasizes the importance of pre-classification, especially in multiple target cases.
- 1 October 2017
- Published by Acoustical Society of America (ASA) in The Journal of the Acoustical Society of America
- Vol. 142 (4_Suppleme) , 2686
Underwater ship noise pattern detection and identification
The anthropogenic underwater noise event could be identified by spectral analysis and acoustic pattern pre-classification of previously measured noise sources. This paper discusses real-time applicable methods for separation of purely natural noise recordings when ships are absent from the data polluted by the ship noise. The natural noise periods are used for the statistical modeling of underwater channel environment and relative levels of marine ambient noise (Wenz curves). The anthropogenic noise periods allow us to identify the noise sources. Moreover, most of the ship identification methods require spectrum component processing and pre-determined ship classification, which could be based on this automatic separation results. The correlational and multiple criteria detection methods compared with cyclostationary feature detection and widely used energy detection. False alarm and misclassification minimization could be achieved by multiple sensor data fusion: weather and Automatic Identification System (AIS). The measured data were collected near the seafloor with an autonomous long-term passive acoustic recorder and were combined with ship passage information from the AIS. Three years of measurements include multiple individual ship observations with different speeds and directions. Analysis of the experimental results emphasizes the importance of pre-classification, especially in multiple target cases.