Multi-Task Learning for Joint Indoor Localization and Blind Channel Estimation in OFDM Systems
Abstract
1. Introduction
2. Related Work
- It uses the raw received signals as input to the localization pipeline.
- It eliminates the need for pilot signals through blind estimation, thereby improving spectral efficiency and reducing the overhead associated with traditional pilot-based methods.
- The multi-task learning framework reduces computational overhead by jointly optimizing channel estimation and localization within the same model, making the system more efficient in terms of both computation and energy consumption.
- The proposed approach achieves superior localization accuracy in both simulated 5G environments (such as NYUSIM) and real-world WiFi CSI environments, demonstrating its effectiveness across different scenarios.
3. Problem Formulation
3.1. Indoor Localization
- ;
- M is the total number of RPs;
- N is the total number of access points;
- K is the number of subcarriers in each CSI measurement.
- M is the total number of RPs;
- Each represents the 2D coordinates (with and being the x- and y-coordinates of the m-th fingerprint location).
3.2. Channel Estimation
3.2.1. Data-Assisted Channel Estimation
- the received symbol at subcarrier k;
- the transmitted symbol at subcarrier k;
- the k-th element of the channel frequency response;
- additive white Gaussian noise.
- be the vector of known transmitted pilot symbols at pilot subcarriers;
- be the corresponding received symbols;
- be the unknown channel frequency response values at the pilot subcarriers;
3.2.2. Blind Channel Estimation
4. Materials and Methods
4.1. Proposed Solution
- represents the -th received symbol at subcarrier k;
- is the transmitted symbol from the first transmitter for the -th symbol at subcarrier k;
- is the transmitted symbol from the second transmitter for the -th symbol at subcarrier k;
- is the CFR for the first transmitter at subcarrier k;
- is the CFR for the second transmitter at subcarrier k;
- is the noise associated with the -th symbol at subcarrier k.
Algorithm 1: Preprocessing mixed channels. |
|
Algorithm 2: Sorting algorithm used in [20]. |
|
- Predict the position of the user terminal using uncertain channels as input.
- Denoise and untangle the channel frequency response pairs to estimate the channel.
- A single-task channel estimation model: an encoder–decoder model made up of a bidirectional LSTM encoder and an LSTM decoder. The encoder processes the multivariate input sequence using four layers of Bi-LSTMs with dropout for regularization. The decoder, also made up of four LSTM layers with dropout, processes the latent representation to generate the sorted and denoised estimations of and .
- A single-task localization model: an encoder–decoder model made up of a Bi-LSTM encoder followed by three fully connected layers predicting the receiver’s coordinates.
- A multi-task model that builds on the single-task models by incorporating a fully connected localization head to the encoder–decoder architecture. For the localization task, three fully connected layers take as input the final hidden state of the decoder in order to predict the corresponding x and y coordinates of the UT. The objective of the multi-task learning approach is for the model to learn a shared representation of the data that is more expressive due to the jointly learned tasks and consequently improve the generalization capabilities of the model as well as the efficiency and performance in comparison to multiple specific models trained to solve a unique task [6].
- Encoder: A 4-layer bidirectional LSTM, processing the input CSI sequence. Each layer applies dropout at a rate of 0.2 for regularization.
- Decoder: A 4-layer unidirectional LSTM, which refines the encoded features for CSI denoising and detangling.
- Channel estimation head: A linear layer maps the decoder output at each subcarrier index to four denoised and detangled CFR output vectors, for H and G, with their respective real and imaginary parts.
- Localization head: The final hidden state from the decoder is passed through three fully connected layers. The final layer predicts the receiver’s position by outputting the 2D coordinates (x, y).
4.2. Experimental Setup
4.2.1. NYUSIM Dataset
4.2.2. WiFi CSI Dataset
5. Results
5.1. Localization Results
5.2. Blind Channel Estimation Results
- The baseline presented in [20], denoted as the initial solution.
- A single-task model using the same encoder–decoder architecture depicted in Figure 3, without the fully connected layers, specifically trained to perform the task of disentangling and denoising and .
- The proposed multi-task solution trained to estimate and as well as predict the location of the receiver.
6. Discussion
Performance Analysis
7. Conclusions
Author Contributions
Funding
Institutional Review Board Statement
Informed Consent Statement
Data Availability Statement
Conflicts of Interest
References
- Puricer, P.; Kovar, P. Technical Limitations of GNSS Receivers in Indoor Positioning. In Proceedings of the 2007 17th International Conference Radioelektronika, Brno, Czech Republic, 24–25 April 2007; pp. 1–5. [Google Scholar] [CrossRef]
- Obeidat, H.; Shuaieb, W.; Obeidat, O.; Abd-Alhameed, R. A Review of Indoor Localization Techniques and Wireless Technologies. Wirel. Pers. Commun. Int. J. 2021, 119, 289–327. [Google Scholar] [CrossRef]
- Bahl, P.; Padmanabhan, V. RADAR: An RF-based user location and tracking system. In Proceedings of the IEEE INFOCOM, Tel Aviv, Israel, 26–30 March 2000; pp. 775–784. [Google Scholar] [CrossRef]
- Chen, L.; Ahriz, I.; Le Ruyet, D. CSI-Based Probabilistic Indoor Position Determination: An Entropy Solution. IEEE Access 2019, 7, 170048–170061. [Google Scholar] [CrossRef]
- Soltani, M.; Pourahmadi, V.; Mirzaei, A.; Sheikhzadeh, H. Deep Learning-Based Channel Estimation. IEEE Commun. Lett. 2019, 23, 652–655. [Google Scholar] [CrossRef]
- Zhang, Y.; Yang, Q. A Survey on Multi-Task Learning. IEEE Trans. Knowl. Data Eng. 2022, 34, 5586–5609. [Google Scholar] [CrossRef]
- Ju, S.; Kanhere, O.; Xing, Y.; Rappaport, T.S. A Millimeter-Wave Channel Simulator NYUSIM with Spatial Consistency and Human Blockage. In Proceedings of the 2019 IEEE Global Communications Conference (GLOBECOM), Waikoloa, HI, USA, 9–13 December 2019; pp. 1–6. [Google Scholar] [CrossRef]
- Munier, F.; Guo, Y.; Da, R. TR 38.857: Study on NR Positioning Enhancements; 3rd Generation Partnership Project (3GPP), Technical Specification Group Radio Access Network, Release 17. 2021. Available online: https://portal.3gpp.org/desktopmodules/Specifications/SpecificationDetails.aspx?specificationId=3732 (accessed on 27 June 2025).
- Yassin, A.; Nasser, Y.; Awad, M.; Al-Dubai, A.; Liu, R.; Yuen, C.; Raulefs, R.; Aboutanios, E. Recent Advances in Indoor Localization: A Survey on Theoretical Approaches and Applications. IEEE Commun. Surv. Tutor. 2017, 19, 1327–1346. [Google Scholar] [CrossRef]
- Wang, X.; Gao, L.; Mao, S.; Pandey, S. CSI-Based Indoor Localization. IEEE Trans. Mob. Comput. 2017, 16, 3251–3264. [Google Scholar]
- Wang, X.; Gao, L.; Mao, S.; Pandey, S. DeepFi: Deep learning for indoor fingerprinting using channel state information. In Proceedings of the 2015 IEEE Wireless Communications and Networking Conference (WCNC), New Orleans, LA, USA, 9–12 March 2015; pp. 1666–1671. [Google Scholar] [CrossRef]
- Chen, Z.; Zhu, Q.; Jiang, H.; Soh, Y.C. Indoor localization using smartphone sensors and iBeacons. In Proceedings of the 2015 IEEE 10th Conference on Industrial Electronics and Applications (ICIEA), Auckland, New Zealand, 15–17 June 2015; pp. 1723–1728. [Google Scholar] [CrossRef]
- Liu, H.; Gan, Y.; Yang, J.; Sidhom, S.; Wang, Y.; Chen, Y.; Ye, F. Push the limit of WiFi based localization for smartphones. In Proceedings of the 18th Annual International Conference on Mobile Computing and Networking, Istanbul, Turkey, 22–26 August 2012; pp. 305–316. [Google Scholar] [CrossRef]
- Zhang, R.; Chen, G.; Cheng, L.; Guan, X.; Wu, Q.; Wu, W.; Zhang, R. Tensor-based Channel Estimation for Extremely Large-Scale MIMO-OFDM with Dynamic Metasurface Antennas. IEEE Trans. Wirel. Commun. Early Access. 2025. [Google Scholar] [CrossRef]
- Li, Y.; Yang, J.; Shih, S.L.; Shih, W.T.; Wen, C.K.; Jin, S. Efficient IoT Devices Localization Through Wi-Fi CSI Feature Fusion and Anomaly Detection. IEEE Internet Things J. 2024, 11, 39306–39322. [Google Scholar] [CrossRef]
- Zhang, B.; Sifaou, H.; Li, G.Y. CSI-Fingerprinting Indoor Localization via Attention-Augmented Residual Convolutional Neural Network. IEEE Trans. Wirel. Commun. 2023, 22, 5583–5597. [Google Scholar] [CrossRef]
- Wu, Z.; Jiang, L.; Jiang, Z.; Chen, B.; Liu, K.; Xuan, Q.; Xiang, Y. Accurate Indoor Localization Based on CSI and Visibility Graph. Sensors 2018, 18, 2549. [Google Scholar] [CrossRef] [PubMed]
- Sánchez-Rodríguez, D.; Quintana-Suárez, M.A.; Alonso-González, I.; Ley-Bosch, C.; Sánchez-Medina, J.J. Fusion of Channel State Information and Received Signal Strength for Indoor Localization Using a Single Access Point. Remote Sens. 2020, 12, 1995. [Google Scholar] [CrossRef]
- Yang, Z.; Zhou, Z.; Liu, Y. From RSSI to CSI: Indoor localization via channel response. ACM Comput. Surv. 2013, 46, 25. [Google Scholar] [CrossRef]
- Terré, M.; Féty, L. Double Blind Channels Estimation for Interference Cancellation of Multicarrier Transmissions. In Proceedings of the 2023 IEEE 34th Annual International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC), Toronto, ON, Canada, 5–8 September 2023; pp. 1–6, ISSN 2166-9589. [Google Scholar] [CrossRef]
- Schuster, M.; Paliwal, K. Bidirectional recurrent neural networks. IEEE Trans. Signal Process. 1997, 45, 2673–2681. [Google Scholar] [CrossRef]
- Ruan, Y.; Chen, L.; Zhou, X.; Liu, Z.; Liu, X.; Guo, G.; Chen, R. iPos-5G: Indoor Positioning via Commercial 5G NR CSI. IEEE Internet Things J. 2023, 10, 8718–8733. [Google Scholar] [CrossRef]
Error (m) | Single-Task | Multi-Task | KNN | DNN | iPos |
---|---|---|---|---|---|
Min | 0.33 | 0.09 | 0.01 | 0.15 | 0.01 |
Mean | 2.76 | 2.40 | 4.58 | 3.69 | 3.00 |
Median | 1.96 | 1.62 | 4.49 | 3.18 | 2.59 |
90th percentile | 5.26 | 5.19 | 8.46 | 9.22 | 6.31 |
Error (m) | Single-Task | Multi-Task | KNN | DNN | iPos |
---|---|---|---|---|---|
Min | 0.11 | 0.15 | 0.01 | 0.15 | 0.00 |
Mean | 2.70 | 2.49 | 4.09 | 3.69 | 3.16 |
Median | 2.59 | 2.27 | 3.57 | 3.18 | 2.87 |
90th percentile | 5.60 | 4.61 | 7.59 | 6.57 | 6.05 |
Error (m) | 3-Tap Channel | 5-Tap Channel | 10-Tap Channel | 30-Tap Channel |
---|---|---|---|---|
Min | 0.09 | 0.21 | 0.07 | 0.04 |
Mean | 2.40 | 2.02 | 1.31 | 0.79 |
Median | 1.62 | 1.24 | 0.89 | 0.58 |
90th percentile | 5.19 | 4.02 | 2.91 | 1.57 |
Method | Inference Time (ms/sample) |
---|---|
Baseline | 33.90 |
Single-task model | 8.51 |
Multi-task model (proposed) | 8.40 |
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2025 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).
Share and Cite
Molina, M.C.; Ahriz, I.; Zerioul, L.; Terré, M. Multi-Task Learning for Joint Indoor Localization and Blind Channel Estimation in OFDM Systems. Sensors 2025, 25, 4095. https://doi.org/10.3390/s25134095
Molina MC, Ahriz I, Zerioul L, Terré M. Multi-Task Learning for Joint Indoor Localization and Blind Channel Estimation in OFDM Systems. Sensors. 2025; 25(13):4095. https://doi.org/10.3390/s25134095
Chicago/Turabian StyleMolina, Maria Camila, Iness Ahriz, Lounis Zerioul, and Michel Terré. 2025. "Multi-Task Learning for Joint Indoor Localization and Blind Channel Estimation in OFDM Systems" Sensors 25, no. 13: 4095. https://doi.org/10.3390/s25134095
APA StyleMolina, M. C., Ahriz, I., Zerioul, L., & Terré, M. (2025). Multi-Task Learning for Joint Indoor Localization and Blind Channel Estimation in OFDM Systems. Sensors, 25(13), 4095. https://doi.org/10.3390/s25134095