Distributed Transceiver Design for Decentralized Estimation in Coexisting IoT Networks

• Published in: IEEE internet of things journal 2024-04, Vol.11 (8), p.14901-14913
• Main Authors: Garg, Navneet, Ratnarajah, Tharmalingam, Vakamulla, Venkata Mani, Sellathurai, Mathini
• Format: Article
• Language: English
• Subjects: Algorithms; Cochannel interference; Convergence; Convexity; Distortionless beamforming; distributed precoding; Estimation; Inference; Interference; interference alignment (IA); Internet of Things; internet of Things (IoT); Kuhn-Tucker method; minimum mean squared error (MSE); Power management; Precoding; Signal to noise ratio; Wireless LAN; Wireless networks; Wireless personal area networks; wireless sensor networks
• ISSN: 2327-4662; 2327-4662
• DOI: 10.1109/JIOT.2023.3345802

With the increasing number of applications of Internet of Things (IoT) devices, co-channel interference is unavoidable among coexisting wireless networks, where multiple IoT devices transmit their observations to their respective destinations [access points (APs)]. In this scenario, we present a joint precoding and power allocation solution to minimize the mean squared error (MSE), while satisfying power constraints at individual IoT devices. In this regard, first, the necessary feasibility condition for the joint convexity of the optimization problem is derived, ensuring the global optimum solution. Subsequently, based on the solution, an iterative MSE algorithm is formulated and analyzed for convergence. The expressions for the MSE-based precoder is obtained via solving Karush-Kuhn-Tucker (KKT) conditions. Further analysis shows that the total resulting MSE at APs is limited by the observation signal-to-noise-ratio (SNR). It leads to the inference that in order to avoid the MSE saturation at APs at higher SNRs, the transmit power at IoT devices should be scaled proportional to and less than the observation SNR. Next, we compare the performance of our solution with two classical methods, namely, the minimum variance distortionless precoding (MVDP) and interference alignment (IA) methods, which are modified and enhanced for the given system. Simulations verify the above inference, and the global convergence of the MSE algorithm, with robustness to initializations yielding the better precoders and power allocation as compared to MVDP's and IA's in terms of the averaged total MSE performance.