A Hybrid Machine Learning and Deep Learning Framework for Accurate Indoor Cooperative Positioning in Wireless Sensor Networks

Indoor cooperative localization for Wireless Sensor Networks (WSNs) faces challenges in achieving accurate positioning due to dynamic channel conditions, dominant non-line-of-sight (NLOS) propagation, and imperfect neighbor selection.In this paper, we propose a hybrid Machine Learning (ML) and Deep Learning (DL) framework to considerably improve accuracy, consistency, and flexibility in indoor localization.Significant technical insights of our framework are three main innovations: the first leverages Gated Recurrent Units (GRUs) with learned parameters to obtain adaptive channel estimation, effectively memorizing through time-varying interference and multipath effects rather than relying solely on the fixed reference path loss from the wall model.Secondly, a Convolutional Neural Network (CNN) is employed for Resilient NLOS Mitigation by learning complex signal distributions to improve range estimates in obscured and cluttered environments.Third, a Reinforcement Learning (RL)-based link optimization scheme enables dynamic selection of neighbor nodes based on real-time signal quality and node density.Experiments demonstrate that our proposed integrated framework enhances localization precision by approximately 21-30% in dynamic indoor environments.It also serves as a robust solution for future smart-city and industrial IoT applications, as well as for autonomous indoor applications.