Distributed Direct Localization Suitable for Dense Networks

Abstract

Traditional network localization algorithms contain ranging and localization steps, which have systematic disadvantages. We propose an algorithm dubbed direct particle filter based distributed network localization (DiPNet). A node's location is directly estimated from the received signals, incorporating location uncertainty of neighboring nodes. The propagation effects on DiPNet become insignificant for dense networks, due to the massive-link collective physical layer processing. DiPNet achieves a near-optimal performance with low complexity, which is particularly attractive for realtime dense-network localization.

Self-Aware Swarm Navigation in Autonomous Exploration Missions

“…In practice, numerous localization algorithms have been developed [27], [41], which can be categorized by the location of estimation into centralized [42] and decentralized algorithms [27], [43]- [45]; by the extractable position-related signal features into signal power, carrier phase or symbol delay-based algorithms [17], [18], [46]; by the measurement of abstraction level into direct localization [47]- [49] and a two-stage approach [17]; by the model of unknown parameters into non-Bayesian algorithms for deterministic parameters like least-square (LS), Gauss-Newton algorithm [46], convex-relaxationbased approaches such as semi-definite programming (SDP) [50] and alternating direction method of multipliers (ADMM) [42], or Bayesian algorithms for random variables [51] like Kalman filter (KF) [45], particle filter (PF) [52], [53], or message passing (MP) algorithms [27], [48], [54].…”

“…459 and 473], and are therefore omitted in this article. For a practical implementation of a swarm navigation system, a decentralized tracking algorithm, see [27], [36], [41], [45], and [48], could be preferable. We first investigate the impacts of the additional RF source and the knowledge of nuisance parameters on the emerging swarm formation.…”

“…In addition, some effects like the RF multipath and NLOS propagation can be modeled stochastically. These effects may be averaged out by collective observations of a swarm [48]. For the dynamic model of an agent, a high-level way point model is applied in this article.…”

Intelligent Agent Technology for Cellular-Assisted GPS Positioning using Bayesian and Self-Organizing Map

“…Various techniques have been introduced over the past decade to improve the accuracy and effectiveness of cellular-based positioning [2]- [5]. However, the reported techniques failed to address the fundamental issue of low accuracy of cellular-based positioning in the absence of GPS and Wireless Local Area Networks.…”

“…Meanwhile the evidence factor Pr(B), which is also known as the normalizing constant, can be viewed as merely a scale factor that guarantees the posterior probabilities are summed to one, as all good probabilities must [8]. Informally, Bayes' formula can be paraphrased as shown in (2).…”