Neighbor Discovery for Wireless Networks via Compressed Sensing

This paper studies the problem of neighbor discovery in wireless networks, namely, each node wishes to discover and identify the network interface addresses (NIAs) of those nodes within a single hop. A novel paradigm, called compressed neighbor discovery is proposed, which enables all nodes to simultaneously discover their respective neighborhoods with a single frame of transmission, which is typically of a few thousand symbol epochs. The key technique is to assign each node a unique on-off signature and let all nodes simultaneously transmit their signatures. Despite that the radios are half-duplex, each node observes a superposition of its neighbors’ signatures (partially) through its own off-slots. To identify its neighbors out of a large network address space, each node solves a compressed sensing (or sparse recovery) problem. Two practical schemes are studied. The first employs random on-off signatures, and each node discovers its neighbors using a noncoherent detection algorithm based on group testing. The second scheme uses on-off signatures based on a deterministic second-order Reed-Muller code, and applies a chirp decoding algorithm. The second scheme needs much lower signal-to-noise ratio (SNR) to achieve the same error performance. The complexity of the chirp decoding algorithm is sub-linear, so that it is in principle scalable to networks with billions of nodes with 48-bit IEEE 802.11 MAC addresses. The compressed neighbor discovery schemes are much more efficient than conventional random-access discovery, where nodes have to retransmit over many frames with random delays to be successfully discovered.

💡 Research Summary

This paper addresses the challenge of neighbor discovery in wireless networks, where each node aims to discover and identify the network interface addresses (NIAs) of nodes within a single hop. A new paradigm called compressed neighbor discovery is proposed, enabling all nodes to simultaneously discover their respective neighborhoods with just one transmission frame, typically consisting of several thousand symbol epochs.

The core technique involves assigning each node a unique on-off signature and allowing simultaneous transmission of these signatures by all nodes. Despite the radios being half-duplex, each node can observe a superposition of its neighbors’ signatures (partially) through its own off-slots. To identify its neighbors from a large network address space, each node solves a compressed sensing or sparse recovery problem.

The paper explores two practical schemes. The first uses random on-off signatures and employs a noncoherent detection algorithm based on group testing for neighbor discovery. The second scheme utilizes on-off signatures based on deterministic second-order Reed-Muller codes and applies a chirp decoding algorithm. This second approach requires significantly lower signal-to-noise ratio (SNR) to achieve the same error performance.

The complexity of the chirp decoding algorithm is sub-linear, making it scalable in principle to networks with billions of nodes using 48-bit IEEE 802.11 MAC addresses. The compressed neighbor discovery schemes are much more efficient than conventional random-access discovery methods, where nodes must retransmit over multiple frames with random delays to be successfully discovered.