Tilbake til søkeresultatene

FRINATEK-Fri prosj.st. mat.,naturv.,tek

Wireless Social Devices Enabled by Self-Organized Spectrum Cartography

Alternativ tittel: WISECAT

Tildelt: kr 12,1 mill.

Med den viktigste utviklingen innen applikasjoner av trådløs teknologi, eksploderende trådløs trafikk og enhetsmangfold, krever fremtidige generasjons mobile trådløse kommunikasjonssystemer høyere datahastigheter og en mer pålitelig overføring samtidig som den opprettholder garantert tjenestekvalitet. Dette møter igjen mange begrensninger i konvensjonelle teknologiske komponenter for å håndtere slik vekst. Følgelig er banebrytende fremskritt nødvendig for å takle de forutsette fremtidige behovene til trådløse systemer. I denne sammenheng gir WISECART et paradigmeskifte for å gjøre det mulig for trådløse heterogene og sosialt engasjerte enheter å danne svært adaptive og selvorganiserte mesh-formasjoner, der kollektiv sosial intelligens introduseres på tvers av alle nettverkslagene. Denne arkitekturen gir høyere grad av fortetting og følgelig bedre spektrumgjenbruk ved å lage en mesh-substrat enhet-til-enhet (D2D) kommunikasjon. Med dette mesh-underlaget utveksles data direkte mellom enheter og radiotilgangspunkter omgås. Videre tildeles ressurser opportunistisk ved å bruke selvorganiserte spektrumkart som lages ved at enhetene selv utnytter deres sanseevne. Vi utforsker datadrevne maskinlæringsalgoritmer for å estimere spektrumkart, og overføre læringsmetoder for å minimere mengden nødvendige sensormålinger. Tilsvarende designer vi også online optimaliseringsalgoritmer og forsterkningslæringsbaserte algoritmer for sekvensiell allokering av nettverksressurser. Brukssakene til WISECART-paradigmet kan være kommunikasjon mellom en samling av mennesker, mellom en gruppe internettkoblede fysiske enheter (Internet of Things - IoT) enheter eller mer generelle integrerte nettkoblede fysiske systemer (Cyber-Physical Systems - CPS). For menneskelige individer er det brukstilfeller som involverer innholdsdeling og distribuert intelligens, inkludert ad-hoc offentlig sikkerhetskommunikasjon og applikasjoner som interaktiv e-læring og e-sport. Det finnes også brukstilfeller for kommunikasjon med mennesker uten innholdsdeling, for eksempel overfylte arenaer, der WISECART-paradigmet brukes til å laste ned trafikk og forbedre energieffektiviteten. I IoT- eller CPS-domenene involverer brukstilfeller intelligente samarbeidende enheter i autonome kjøretøy og e-helse.

Main outcomes: i) design of spectrum cartography algorithms for both spectrum maps and channel gain maps ii) design of computationally efficient spectrum cartography algorithms that perform well in the absence of accurate location information. iii) design of spectrum cartography algorithms that perform well in cases of data scarcity and sparsity. iv) design of computationally efficient resource allocation algorithms for next-generation wireless networks: v) design of robust resource allocation algorithms that can perform well using imperfect CSI estimates. vi) design of data-driven coding and detection algorithms that are robust and computationally efficient. vii) design of end-to-end machine learning algorithms for resource allocation and detection in various wireless communication networks. Scientific and industrial impact: From a research point of view, we have designed novel algorithms for a) resource allocation for D2D communications, b) spectrum cartography for radio maps and channel gains, c) resource allocation in NOMA communications, d) joint CSI estimation and detection for non-orthogonal multiple access (NOMA) communication. We have developed high expertise in data-driven model learning, online optimization, and machine learning for problems related to next-generation wireless communications. In addition, a large percentage of this expertise is also beneficial to find solutions to problems in other application domains involving resource allocation and estimation. In addition, some of the results obtained in WISECART have contributed also toward paving the way to the next generation wireless networks, including 5G and 6G. As a consequence, all the graduated PhD students have been hired (some times even before graduation) at highly internationally recognized institutions, such as NOKIA Bell Labs, Ericsson Research and SIMULA, among others. We have also established close relations with several Indian Institutes of Technology in connection to the WISECART project, enforcing collaborations on several follow-up topics. We continue to disseminate results from the WISECART project in these institutions and analyzing possible transfer of the results in WISECART to other related research problems. Educational impact: We have also integrated some of the activities and research results of the project into some advanced Master or PhD courses: Deep Learning and Deep Reinforcement Learning, which are directly related to the topics covered in this project, have been also open to students from other universities. We plan to design mini-projects inside these courses that include data and problems from real networks, in cooperation with some of the industrial partners that have hired WISECART graduated PhD students, such as NOKIA Bell Labs, Ericsson Research or SIMULA.

WISECART proposes a drastically new concept for the design of future wireless heterogeneous networks, including both mobile terminals and different types of devices in the context of Wireless Sensor Networks and the emerging paradigms of Internet-of-Things and Cyber-physical Systems, in order to meet the extremely demanding future societal needs. The main idea is to provide wireless heterogeneous networks and devices with the autonomous capability to form highly adaptive collectively social intelligent local mesh-type formations, by means also of novel machine learning algorithms. The operational parameters of the network (from physical to network layer) are adjusted according to their interaction with the environment, energy availability and, most importantly, on the communication needs established by either a user or a collection of socially engaged users and devices. The new concept of Club, along with Spectrum Cartography, makes this research ground-breaking. This requires wireless devices and networks to have the following challenging features (not currently present today): a) endowment with some native social intelligence across all the communication layers, not only at the application layer as in current solutions, b) full capability to coordinate locally, in a self-organized manner, the space-time-frequency radio resources, providing high flexibility,robustness and reactivity to dynamic rapid changes in local radio activity and heterogeneous traffic demands; c) much higher degree of decentralization for coordination mechanisms in these networks, which in turn requires also a joint energy-aware optimization of the distributed sensing, computation and communication tasks across the nodes, instead of treating them as separate optimization problems, as in previous work. Experimental evaluation will be performed through both simulations and by constructing a wireless network demonstrator composed of heterogeneous terminals with sensing, cognitive and reactive capabilities.

Publikasjoner hentet fra Cristin

Ingen publikasjoner funnet

Ingen publikasjoner funnet

Ingen publikasjoner funnet

Budsjettformål:

FRINATEK-Fri prosj.st. mat.,naturv.,tek