Fast Neighbor Discovery in MEMS FSO Networks

Abstract

We investigate whether it is possible to achieve sub-millisecond latency for the discovery of multiple neighbors in laser-based Free Space Optical (FSO) networks. Given a large programmable array of micromirrors, we propose the use of adaptive boolean combinatorial group testing algorithms that are practical and efficient. The time taken scales as O(N log(L/N)) for N neighbors even if N is unknown, but no additional computation (e.g., matrix inversion) is required. Compared to Raster and Lissajous pattern-based scanning, we report 99.92% and 87% reduction in latency, respectively, for an array of 106 micromirrors (approximately XGA resolution). We conclude that it is indeed possible to achieve sub-ms latency given realistic network parameters. Our proposed algorithms are evaluated in simulation, and are compared against state of art neighbor discovery schemes.