[Paper Review] Distributed Ranging and Localization for Wireless Networks via Compressed Sensing

Ming Gan, Dongning Guo+1 more|arXiv (Cornell University)|Aug 16, 2013

Indoor and Outdoor Localization Technologies21 references3 citations

TL;DR

This paper proposes a compressed sensing-based physical-layer framework for distributed ranging and localization in wireless networks, enabling simultaneous multi-node transmission and reception via on-off signaling to overcome half-duplex constraints. Simulations show convergence to sub-meter accuracy within ~10 iterations using 1,200-symbol frames, significantly outperforming contention-based protocols like ALOHA or CSMA.

ABSTRACT

Location-based services in a wireless network require nodes to know their locations accurately. Conventional solutions rely on contention-based medium access, where only one node can successfully transmit at any time in any neighborhood. In this paper, a novel, complete, distributed ranging and localization solution is proposed, which let all nodes in the network broadcast their location estimates and measure distances to all neighbors simultaneously. An on-off signaling is designed to overcome the physical half-duplex constraint. In each iteration, all nodes transmit simultaneously, each broadcasting codewords describing the current location estimate. From the superposed signals from all neighbors, each node decodes their neighbors' locations and also estimates their distances using the signal strengths. The node then broadcasts its improved location estimates in the subsequent iteration. Simulations demonstrate accurate localization throughout a large network over a few thousand symbol intervals, suggesting much higher efficiency than conventional schemes based on ALOHA or CSMA.

Motivation & Objective

• To address the inefficiency of contention-based medium access (e.g., ALOHA, CSMA) in conventional ranging and localization protocols.
• To enable simultaneous location estimation and distance measurement among all neighbors in a single transmission frame.
• To overcome the physical half-duplex constraint in wireless nodes using rapid on-off division duplex (RODD) signaling.
• To design a fully distributed, iterative algorithm that converges to accurate node location estimates using only local information exchange and compressed sensing.
• To demonstrate that physical-layer integration of ranging and localization yields significantly higher spectral and time efficiency than traditional network-layer solutions.

Proposed method

• Each node broadcasts codewords encoding its current location estimate during designated on-slots in a synchronized frame.
• Nodes listen during off-slots to receive superposed signals from all neighbors, leveraging the superposition property of wireless channels.
• Using compressed sensing, each node decodes the location estimates and signal strengths (for ranging) from the superimposed signals.
• Signal strength measurements are used to estimate distances to neighbors, assuming path loss models.
• The RODD signaling scheme enables full-duplex-like operation by alternating transmission and reception within the same frame.
• Each node iteratively updates its location estimate using decoded neighbor information and re-broadcasts the improved estimate in the next frame.

Experimental results

Research questions

• RQ1 Can simultaneous multi-node communication in wireless networks be leveraged for joint ranging and localization at the physical layer?
• RQ2 How can the physical half-duplex constraint be overcome to allow concurrent transmission and reception in distributed localization?
• RQ3 To what extent can compressed sensing techniques enable accurate location recovery from superimposed, noisy signals in a distributed network setting?
• RQ4 How does the convergence speed and localization accuracy scale with network density, anchor distribution, and SNR?
• RQ5 Can a fully distributed, iterative algorithm achieve high localization accuracy without centralized coordination or prior synchronization beyond frame-level timing?

Key findings

• The proposed algorithm achieves sub-meter localization accuracy (typically a small fraction of a meter) for nodes within the convex hull of their neighbors, even in large networks with 100 nodes.
• In a 100-node network with 16 anchors in a 4×4 lattice, the average localization error decreases monotonically over ~10 iterations, reaching a stable low error level.
• Over 90% of clients achieve localization error within 1 meter after 8 iterations, with performance improving further in later iterations.
• The average localization error drops sharply with increasing SNR below 30 dB, but plateaus above 30 dB due to interference dominance.
• The scheme requires only ~10 iterations of 1,200-symbol frames, making it significantly more efficient than contention-based protocols that require many retransmissions due to collisions.