Consensus-based Algorithms for Distributed Network-State Estimation and Localization

Recent advances of hardware design and radio technologies have opened the way for an emerging category of network-enabled smart physical devices as a result of convergence in computing and wireless communication capabilities. Inspired by biological interactions, distributed processing of data collected by individual devices is now becoming crucial to let the nodes self-learn relevant network-state information and self-organize without the support of a central unit. Focus of this paper is twofold. First, a novel network channel model tailored for dense deployments is developed and validated on real data. The model describes relevant channel features that are representative of site-specific static/dynamic multipath fading and are shared by all links of a network. Second, a new class of distributed weighted-consensus strategies is introduced to support distributed network calibration and localization in device-to-device networks. Network calibration allows the devices to self-learn the common channel parameters, by successive refinements of local estimates and peer-to-peer information exchange. Network-localization enables each node to acquire augmented information about the whole network topology, by distributed learning from local channel observations. The proposed distributed algorithms guarantee a fast convergence and can replace conventional centralized schemes. An experimental case study is discussed in a representative indoor environment for the purpose of system validation. Experimental results show that the proposed method can significantly improve the performance of conventional solutions.