An efficient and optimized cooperative indoor localization for B5G communication

Deepti Sharma,Ramesh Babu Battula
DOI: https://doi.org/10.1002/dac.5702
2024-01-20
International Journal of Communication Systems
Abstract:This paper introduces an innovative MOIL framework that utilizes deep learning‐based localization for B5G communication. The objective was to provide highly precise positioning information for user equipment. The proposed framework comprises two functional blocks: SecLOP and CordNet. SecLOP incorporates BiLSTM as a building cell that extracts and enhances signal information from nearby access points by mitigating multipath effects and noise. CordNet uses fully connected dense layers considering all signal features combination and providing the final geometric location of the user. Summary Ultra‐accurate indoor localization/positioning technology is rapidly gaining interest in 5G and beyond (B5G) communications. Achieving 0.1 m level of precise indoor localization is a significant challenge that can be resolved using channel state information (CSI) of wireless signals from nearby access points. The complex indoor propagation issues like shadowing, blockage effects, and multi‐path fading diminish CSI, affecting precise position/location accuracy. Thus, to achieve ultra‐accurate localization, a novel MOIL framework is proposed using a deep learning BiLSTM model and CSI features in an efficient and optimized manner. The proposed MOIL integrates two functional blocks, SecLOP and CordNet. SecLOP conceals multi‐path effects and noise to optimize the input information using sequential responses of signals. CordNet minimizes precise position/location errors by taking all combinations of optimized signal features from the SecLOP block. Extensive simulation experiments were conducted to validate MOIL in different indoor scenarios, such as line‐of‐sight (LOS) and non‐line‐of‐sight (NLOS). Moreover, through a substantial ablation study, the efficiency of the proposed framework is validated/verified. The results outperformed existing works by providing cm‐level accuracy of 1 cm in LOS and 1.5 cm in NLOS scenarios. The proposed work outperformed the literature by 31.3% in LOS and 25.30% in NLOS scenarios.
telecommunications,engineering, electrical & electronic
What problem does this paper attempt to address?