A Meta-learning based Generalizable Indoor Localization Model using Channel State Information

• DOI: 10.48550/arxiv.2305.13453
• Publication Date: 2023-01-01
• Authors: Ali Owfi; ChunChih Lin; Linke Guo; Fatemeh Afghah; Jonathan Ashdown; Kurt Turck
• Keywords: Retraining; Channel state information

Indoor localization has gained significant attention in recent years due to its various applications smart homes, industrial automation, and healthcare, especially since more people rely on their wireless devices for location-based services. Deep learning-based solutions have shown promising results accurately estimating the position of indoor environments using parameters such as Channel State Information (CSI) Received Signal Strength Indicator (RSSI). However, despite success deep approaches achieving high accuracy, these models suffer from a lack generalizability can not be readily-deployed new or operate dynamic without retraining. In this paper, we propose meta-learning-based address that persists conventionally trained DL-based models. Furthermore, meta-learning algorithms require diverse datasets several different scenarios, which hard collect context localization, design algorithm, TB-MAML (Task Biased Model Agnostic Meta Learning), intended further improve when dataset is limited. Lastly, evaluate performance TB-MAML-based against done other algorithms.