A Transfer Learning-Driven Methodology for Efficient and Sustainable Indoor Localization Using VLC

Masood Jan; Wafa Njima; Xun Zhang; Alexander Artemenko

cs.LGOct 21, 2026

A Transfer Learning-Driven Methodology for Efficient and Sustainable Indoor Localization Using VLC

Authors: Masood Jan, Wafa Njima, Xun Zhang, Alexander Artemenko

Organizations: Doctoral School of Computer Science, Telecommunications and Electronics (EDITE), Sorbonne University, Paris, France · LISITE Laboratory, Institut Supérieur d’Électronique de Paris, Paris, France · Université Paris-Saclay, LISV Laboratory, France · the sector research of Robert Bosch GmbH in Stuttgart, Germany

Abstract

The rapid expansion of the Internet of Things enables seamless collaboration among connected devices, making indoor localization a critical component of industrial automation. However, conventional localization techniques struggle in dynamic environments. These methods depend on extensive, frequent data collection and environment-specific calibration, limiting scalability, interoperability, and effective use of prior research while imposing significant energy penalties that hinder practical deployment. To address these challenges, we propose an energy-efficient transfer learning (TL) based approach for visible light communication based indoor localization. Our method leverages TL to tackle environmental variability, including lighting fluctuations and physical obstacles that typically degrade localization performance, while simultaneously reducing computational overhead and energy consumption compared to conventional approaches. We also introduce a novel model efficiency (ME) metric, designed to integrate localization accuracy, energy efficiency, data efficiency, and transfer gain into a single evaluative measure for comprehensive optimization. Evaluations on a real-world dataset collected from a BOSCH factory demonstrate that our model achieves a 47% improvement in localization accuracy compared to a conventional model that does not utilize TL, and achieves up to 67.5% ME with TL models, compared to only 10.4% in conventional models under high-noise conditions. These results underscore the potential of our approach to deliver a highly efficient and scalable indoor localization system suitable for energy-constrained industrial applications.

Figures & tables

Fig. 1 : System model illustrating the TL-based framework for model development and evaluation.

Fig. 2 : Proposed TL-based localization process.

Fig. 3 : Architecture of the implemented VLC transceiver.

Category	Parameter	Value
Power Supply	Supply voltage	12 V
	Supply current	1.2 A
	Power supply interface	USB-C female
TX (Transmitter)	Driving current signal range	200 mA $\pm$ 100 mA
	Transmitted optical power range	500 mW $\pm$ 250 mW
	Full angle at half power	60 ∘

TABLE I : VLC Transceiver Specifications.

Fig. 5 : Schematic of the industrial hall at BOSCH in Blaichach.

Panel of Figure 7 — Fig. 6 : Illustrations of (a) Tx/Rx setup, and (b) Floor plan layout.

Transmitter ID	Position (x, y, z)	Located Area
TX1	(2.6, 0.8, 2.9)	Near control station and transparent windows
TX2	(1.7, -0.2, 2.9)	Near control station and transparent windows
TX3	(3.7, 1.8, 2.9)	Near control station and transparent windows
TX4	(17.1, 0.2, 2.9)	Material storage zone
TX5	(12.9, 0.8, 2.9)	Rear of production machinery (beside cement pillars)
TX6	(17.1, 0.8, 2.9)	Rear of production machinery (adjacent to storage)

TABLE II : 3D Coordinates and Located Areas of Transmitters.

Parameter	Description	Value
$d$	Input dimension	10
Optimizer	Model optimizer	ADAM
$\eta$	learning rate	0.0001
$B$	Batch size	64
$E$	Number of epochs per iteration	600
$L$	Number of hidden layers	5

TABLE III : TL Simulation Settings.

Panel of Figure 10 — Fig. 7 : TL performance evaluation: (a) Localization Error (b) Success Rate (c) TL of Base Model with different NFs.

Final Metric Value	NF = 0	NF = 2		NF = 4		NF = 8
Training Model	Base Model	EV Model	TL Model	EV Model	TL Model	EV Model	TL Model
Training Localization Error (cm)	41.13	70.38	41.37	78.99	41.74	79.91	42.42
Validation Localization Error (cm)	41.21	70.17	41.37	79.37	41.71	80.41	42.59
Training Success Rate (%)	93.07	57.76	93.07	47.25	92.32	47.75	92.49
Validation Success Rate (%)	91.64	59.20	91.97	47.16	90.97	47.83	91.30
Cumulative Energy Consumed (J)	852.79	1571.35	1064.91	1923.38	1082.56	976.70	867.22

TABLE IV : Comparison of final metrics across models with different NFs, using the base model at NF = 0.

NF	EV Model (cm)	TL as Base Model (cm)	TL as (EV Model NF =2) (cm)	TL as (EV Model NF =4) (cm)
0	-	41.19	41.17	41.14
2	70.17	41.37	41.25	41.20
4	79.37	41.71	41.43	41.65
8	80.41	42.59	41.26	41.73

TABLE V : Localization error comparison of TL models at different NFs.

Data Size	Error Before TL (cm)			Error After TL (cm)
	NF = 2	NF = 4	NF = 8	NF = 2	NF = 4	NF = 8
30%	77.08	83.08	86.78	41.89	42.05	42.96
50%	76.49	81.89	84.75	41.71	41.65	41.89
70%	74.66	80.17	81.33	41.63	41.45	41.59
100%	70.17	79.01	80.76	41.45	41.32	41.33

TABLE VI : Error values before and after TL at different NFs, comparing performance across various data sizes.

Fig. 8 : TL model evaluation with limited data.

Panel of Figure 15 — Fig. 9 : Key Metrics Evaluation: (a) Transfer Gain for varying data sizes and NFs (b) Performance Accuracy of Base, EV, and TL models (c) Energy Efficiency trends derived from experimental measurements.

Fig. 10 : Model efficiency across different models and environmental conditions under varying data proportions and noise factors.

Panel of Figure 17 — Fig. 11 : (a) CDFs of localization error for Base, EV, and TL models under varying NFs and (b) Direct testing of the best ME model across different NFs. The dashed line indicates the 80th-percentile error threshold (80.1 cm).

Wireless Localization

Wireless localization uses radio signals and sensor fusion to estimate the positions and trajectories of people, devices, and transmitters.

See all 65 papers

Cross-Area Transfer Learning for VLC-based Indoor Localization with a Transfer Efficiency Score
Oct 21, 2026Masood Jan, Wafa Njima, Xun Zhang +1Wireless Localization Cross-Domain Transfer Learning
Doctoral School of Computer Science, Telecommunications and Electronics (EDITE), Sorbonne University, 75006 Paris, France · LISITE Laboratory, Institut Supérieur d’Électronique de Paris, 75006 Paris, France · Université Paris-Saclay, LISV Laboratory, France +1
Deep Attention-based Sequential Ensemble Learning for BLE-Based Indoor Localization in Care Facilities
Mar 22, 2026Minh Triet Pham, Quynh Chi Dang, Le Nhat TanWireless Localization Ensemble Learning
School of Computing and Information Systems, The University of Melbourne, Melbourne, Australia · Faculty of Applied Science, Ho Chi Minh City University of Technology, Vietnam National University - Ho Chi Minh City, Ho Chi Minh City, Vietnam
ARMOR: Adaptive Resilience Against Model Poisoning Attacks in Continual Federated Learning for Mobile Indoor Localization
Mar 20, 2026Danish Gufran, Akhil Singampalli, Sudeep PasrichaFederated Learning Federated Continual Learning
Colorado State University, Fort Collins, CO 80523, USA

Oct 21, 2026J/K move · Enter open · S save

Cross-Area Transfer Learning for VLC-based Indoor Localization with a Transfer Efficiency Score

Masood Jan, Wafa Njima, Xun Zhang, Alexander Artemenko

Doctoral School of Computer Science, Telecommunications and Electronics (EDITE), Sorbonne University, 75006 Paris, France · LISITE Laboratory, Institut Supérieur d’Électronique de Paris, 75006 Paris, France · Université Paris-Saclay, LISV Laboratory, France+1

Wireless Localization Cross-Domain Transfer Learning

Open paper AskPDF arXiv

Indoor localization systems are critical for Industry-4.0 applications. Visible Light Communication (VLC)-based systems offer advantages such as immunity to electromagnetic interference, high accuracy, and energy efficiency, but their performance is hindered by environmental variability across deployment areas. This paper presents a transfer learning (TL) framework for VLC-based indoor localization that enhances cross-area robustness while reducing data requirements. Using real-world measurements from four distinct areas of a Bosch manufacturing facility, we show that fine-tuning a deep neural network (DNN) pre-trained on source area data with only 40% of target-area measurements recovers 97% of full-data performance, thereby reducing the amount of target data needed. We further introduce the Transfer Efficiency Score (TES), a composite metric that identifies the most effective source models for transfer without exhaustive evaluation. Guided by TES, the framework selects the best source model, yielding localization errors as low as 40.9 cm with success rates above 86%. Validation on an unseen area confirms its generalization capability, with a 3.28 cm mean error achieved using the TES-selected source model. Overall, the proposed framework provides a scalable and data-efficient pathway for industrial VLC localization, substantially reducing deployment costs and computational burden while maintaining high accuracy in diverse and dynamic environments.

Abstract

Wireless Localization

Explore similar work

Cross-Area Transfer Learning for VLC-based Indoor Localization with a Transfer Efficiency Score

Deep Attention-based Sequential Ensemble Learning for BLE-Based Indoor Localization in Care Facilities

ARMOR: Adaptive Resilience Against Model Poisoning Attacks in Continual Federated Learning for Mobile Indoor Localization