Deep Learning-Based Data Fusion of 6G Sensing and Inertial Information for Target Positioning: Experimental Validation

Karthik Muthineni 1 3 , Alexander Artemenko 1 , Artjom Grudnitsky 2 , Josep Vidal 3 , Montse Nájar 3 · 2025

Read Deep Learning-Based Data Fusion of 6G Sensing and Inertial Information for Target Positioning: Experimental Validation by Karthik Muthineni 1 3 , Alexander Artemenko 1 , Artjom Grudnitsky 2 , Josep Vidal 3 , Montse Nájar 3 (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-24253.txt.

Introduction

Muthineni and co-authors present a 2025 paper on locating targets by fusing 6G network sensing with motion sensors. Future 6G networks will sense their surroundings like radar while they carry data. A target's own inertial sensors add private detail about its motion. Blending both streams should sharpen position fixes indoors.

First the team gauges how well network sensing alone can place a moving target. Then they add the target's sensor data through a deep learning fusion method. Tests ran on a real proof-of-concept rig, not just in simulation. Gains from fusion show up in the measured accuracy. The work points to tighter indoor tracking for next-generation networks.

Catalog record

CREATORKarthik Muthineni 1 3 , Alexander Artemenko 1 , Artjom Grudnitsky 2 , Josep Vidal 3 , Montse Nájar 3
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE27 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-24253

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Deep Learning-Based Data Fusion of 6G Sensing and Inertial Information for Target Positioning: Experimental Validation Karthik Muthineni 1 3 , Alexander Artemenko 1 , Artjom Grudnitsky 2 , Josep Vidal 3 , Montse Nájar 3 Abstract The sixth-generation (6G) cellular technology will be deployed with a key feature of Integrated Sensing and Communication (ISAC), allowing the cellular network to map the environment through radar sensing on top of providing communication services. In this regard, the entire network can be considered as a sensor with a broader Field of View (FoV) of the environment, assisting in both the positioning of active and detection of passive targets. On the other hand, the non-3GPP sensors available on the target can provide additional information specific to the target that can be beneficially combined with ISAC sensing information to enhance the overall achievable positioning accuracy. In this paper, we first study the performance of the ISAC system in terms of its achievable accuracy in positioning the mobile target in an indoor scenario. Second, we study the performance gain achieved in the ISAC positioning accuracy after fusing the information from the target’s non-3GPP sensors. To this end, we propose a novel data fusion solution based on the deep learning framework to fuse the information from ISAC and non-3GPP sensors. We validate our proposed data fusion and positioning solution with a real-world ISAC Proof-of-Concept (PoC) as the wireless infrastructure, an Automated Guided Vehicle (AGV) as the target, and the Inertial Measurement Unit (IMU) sensor on the target as the non-3GPP sensor. The experimental results show that our proposed solution achieves an average positioning error of 3 ​ cm 3~\textrm{cm} , outperforming the considered baselines. I Introduction The upcoming sixth-generation (6G) mobile communication technology is expected to extend its footprint from the traditional consumer market to vertical industries by promising to deliver advanced functionalities that open new use cases [ 1 ] . The 6G technology envisions taking the network capability beyond communications by enabling “radar-like” sensing to detect objects and their positions without having them equipped with a radio transceiver. This is achieved with a functionality called Integrated Sensing and Communication (ISAC), with integration performed at different stages, such as spectrum sharing between radar sensing and communication functions, use of the same hardware, and reuse of the communication waveform and radio resources for sensing [ 2 , 3 ] . Nevertheless, this integration increases the system’s complexity and opens new challenges. The development of cellular networks relied heavily on foundation models and theories. However, these approaches face limitations due to increased network complexity as we move from generation to generation. Recent years have witnessed the enormous capabilities of Artificial Intelligence (AI) algorithms in solving complex models and optimization problems in various applications [ 4 , 5 ] . The role of AI can also be witnessed across a broad range of communication topics like signal detection [ 6 ] , channel encoding [ 7 ] , channel estimation [ 8 ] , and resource allocation [ 9 ] . Moreover, the future 6G cellular network is expected to natively support AI processing by adding suitable hardware components into the network design [ 10 ] . Therefore, AI will be key in addressing certain challenges of 6G and the ISAC system [ 11 ] . Fig. 1 : Block diagram of our ISAC sensing/positioning setup. The standardization body 3GPP has conducted a feasibility study on ISAC in Release 19 and identified 32 potential use cases [ 12 ] . This paper evaluates the sensing/positioning performance of the ISAC system introduced in [ 13 ] and the potential of AI models in enhancing the ISAC-assisted positioning accuracy for one of the 3GPP’s use cases, i.e., tracking Automated Guided Vehicles (AGVs) in complex scenarios. Earlier works on ISAC-assisted tracking focused on evaluating the ISAC system performance in simulated environments. In [ 14 ] , authors considered multiple mono-static radars for fusing sensing information corresponding to a single target and compared the tracking performance against different algorithms. A joint framework for estimating the target location and its clock drift in the 6G ISAC system was presented in [ 15 ] . In [ 16 ] , the authors proposed using an AI model for positioning the target by fusing the sensing information collected by multi-static radars in an indoor scenario. There is still a research gap in ISAC literature investigating how to fuse ISAC sensing information with the data obtained from other sensor modalities, such as the non-3GPP sensors available on the target. We contribute to ISAC literature by ( i ) evaluating the achievable positioning accuracy of the real mobile target in an indoor scenario using a Proof-of-Concept (PoC) of ISAC, ( ii ) demonstrating the value of non-3GPP sensors and their provided information in enhancing the ISAC-assisted positioning accuracy by fusing the information from these two systems, ( iii ) we propose a novel data-driven Deep Neural Network (DNN) based data fusion approach for fusing the non-3GPP data with ISAC sensing information. Furthermore, we validate our approach through real-world experiments with an ISAC PoC as the wireless infrastructure, the AGV serving as the passive target for positioning, and an Inertial Measurement Unit (IMU) sensor on the target as the non-3GPP sensor. To the best of our knowledge, this is the first study focusing on enhancing ISAC-assisted target positioning accuracy with non-3GPP data and demonstrating the results through real-world experiments. II Background The ISAC PoC used in this work is built upon the 5G communications hardware [ 13 ] . In this section, we present the details of ISAC PoC, the achievable positioning accuracy with ISAC, and the problem under investigation. II-A ISAC System We consider a quasi-monostatic 1 1 1 A configuration where transmitter and receiver are co-located but are slightly separated in distance. ISAC PoC, which uses base station-centric sensing, i.e., transmission and reception of sensing-relevant signals, and their processing does not involve User Equipment (UE). Sensing is performed using the Orthogonal Frequency Division Multiplexing (OFDM) radar algorithm [ 17 ] , where a transmitted OFDM radio frame with \mathrm{G} symbols and \mathrm{H} subcarriers, forming a matrix U , is reflected by the environment. The reflections are received by the ISAC antennas as an OFDM radio frame forming the matrix V , allowing computation of the Channel State Information (CSI) matrix Z by element-wise division of V by U . The OFDM radar algorithm then performs \mathrm{H} Fast Fourier Transforms (FFTs) of length \mathrm{G} and \mathrm{G} Inverse Fast Fourier Transforms (IFFTs) of length \mathrm{H} on the CSI matrix Z , to obtain a range-doppler matrix [ 18 ] . We enhance the OFDM radar algorithm to improve target detection with the clutter removal approach described in [ 19 ] . The magnitude of each element in the range-doppler matrix is used to differentiate between noise and potential targets, with a simple threshold-based approach used in the following experiments. For each target thus obtained, we extract the parameters range r and Doppler velocity v_{\textrm{D}} . The PoC realizing this system consists of 2 Radio Units (RUs): a gNodeB (gNB) RU used for transmission of both sensing and communication signals and an additional RU of the same type used only

Read the complete text