Unsourced Random Access with Correlated Devices
Author
Stern, Kristoffer
Term
4. semester
Education
Publication year
2019
Submitted on
2019-06-06
Pages
133
Abstract
Fremtidens trådløse net skal kunne håndtere enorme mængder af maskiner og sensorer, der kommunikerer uden menneskelig indgriben, fx i internet of things. I analyser og design antages det ofte, at enhederne opfører sig uafhængigt. I mange praktiske IoT-scenarier, som distribuerede sensornetværk, reagerer enhederne dog på den samme fysiske hændelse. Det betyder, at både deres aktivitet og de data, de sender, kan være korreleret. Vi undersøger en model, der eksplicit indfanger denne korrelation. En fælles fysisk begivenhed kan udløse en alarm, så en delmængde af enheder sender den samme besked på samme tid. Vi udvikler en informations-teoretisk model for fejl-sandsynlighed, som ud over manglende detektering også medtager falske positiver – situationer hvor modtageren fejlagtigt afkoder en bestemt besked, selv om ingen enhed sendte den. Vores resultater viser, at det at udnytte korrelation kan give meget høj pålidelighed, men det sker på bekostning af netværkets spektrale effektivitet, dvs. hvor effektivt radiospektret udnyttes. Derudover kan ikke-ortogonal adgang med superpositionskodning – hvor flere enheders signaler overlapper i tid og frekvens og lægges oven på hinanden – være at foretrække frem for ortogonal adgang – hvor enheder adskilles i tid eller frekvens – når flerbrugerinterferens er lav til moderat.
Future wireless networks must connect vast numbers of machines and sensors that communicate without human involvement, for example in the internet of things. Analyses and designs often assume devices act independently. In many real IoT settings, such as distributed sensor networks, devices respond to the same physical event, so both their activity and the data they send can be correlated. We study a model that explicitly captures this correlation. A common physical phenomenon can trigger an alarm, causing a subset of devices to transmit the same message at the same time. We develop an information-theoretic error probability model that, beyond missed detections, also accounts for false positives – cases where the receiver incorrectly decodes a specific message even though no device sent it. Our results show that exploiting correlation can deliver very high reliability, but at the cost of network spectral efficiency, that is, how efficiently the available radio spectrum is used. In addition, non-orthogonal access with superposition encoding – allowing multiple devices’ signals to overlap in time and frequency and be layered – can be preferable to orthogonal access – separating devices in time or frequency – when multi-user interference is low to moderate.
[This abstract was generated with the help of AI]
Keywords
Documents