Global Spectrum Opportunity Assessment WiFiUS Status Report

Transcription

Global Spectrum Opportunity Assessment WiFiUS Status Report
8.2.2016
Global Spectrum Opportunity Assessment
WiFiUS Status Report
Dennis Roberson
10 December 2015
Co-Principle Investigators and Contributors
• Marja Matinmikko – VTT Technical Research Ctr
• Xianfu Chen, Marko Höyhtyä, Aarne Mämmelä
• Jarkko Paavola –Turku University of Applied Science
• Reijo Ekman, Juhani Hallio, Juha Kalliovaara, Jani Auranen
• Juha Roning – University of Oulu
• Jaakko Suutala, Anna-Mari Vimpari, Jaakko Lampi
• Allen MacKenzie – Virginia Tech
• Ramakrishnan Kalyanaraman, Abdallah Abdallah
• Dennis Roberson – Illinois Institute of Technology
• Ryan Attard, Abed Arnaout, Billy Bafia, Roger Bacchus,
Yupeng Dong, Eric Faurie, Yu Gu, Cindy Hood, Sohail
Noor, Marriyam Qureshi, Ali Riaz, Philip Felber, Tanim
Taher, Jesse Taylor, Emilie Woog, Ken Zdunek
1
8.2.2016
Research Achievements vs. Goals
1) Deployed 5 (vs. goal of 3) spectrum observatories in
Finland (2 – 1 fixed and 1 mobile), at Virginia Tech (1) and
at IIT (2+). Observatories based on a common platform
and generating a single spectrum measurement database
2) Developed empirically validated, statistical models of
spectrum utilization for different wireless application types
based on this dataset
3) Mined the data using “big data” analytical techniques to
discover temporal and spectral correlations and
relationships not obvious using traditional approaches
4) Using models to develop temporal / spectral occupancy
predictions for a variety of wireless application categories
5) Collaborators are leveraged existing national, region and
international government regulatory relationships to share
the research results & influence spectrum management
policy with the goal of more efficient spectrum utilization
RFeye based Spectrum Observatory
2
8.2.2016
WiFiUS Spectrum Observation
Turku, Backpack System
Mobile System “Head”,
Chicago
Mobile Systems, Chicago
WiFiUS Spectrum Observatory Sites in Chicago
Harbor Point, Chicago
IIT Tower, Chicago
3
8.2.2016
WiFiUS Spectrum Observatory Sites
TUAS - Turku, Finland
Virginia Tech - Blacksburg, VA
Spectrum Observatory Site – TUAS building
- Sepänkatu 1, Turku, Finland – July 2013
4
8.2.2016
Developed / Deployed Web-based Spectrum Monitor
http://www.cs.iit.edu/~wincomweb/live-monitor.html
Power Spectrum Plot for Frequency range of
1755 – 1850MHz
Collected in Turku, Finland on 17 vs. 26 June, 2013
Average PSD
-50
-60
-80
-90
Average PSD
-100
-50
-110
-60
-120
1760
1770
1780
1790 1800 1810
Frequency (MHz)
1820
1830
1840
Power in dBm
Power in dBm
-70
-70
1850
-80
-90
-100
-110
1760
1770
1780
1790
1800
1810
Frequency (MHz)
1820
1830
1840
1850
5
8.2.2016
Spectrum Observatory
WiFiUS Measurement Band Plan
Band
Freq. range
(MHz)
Resolution
bandwidth
Scan
interval
RF
Input
Port
1
30-130
78.125 kHz
10 s
1
2
130-800
39.0625 kHz
3s
1
3
650-1200
39.0625 kHz
3s
2
4
1200-3000
39.0625 kHz
3s
3
5
3000-6000
78.125 kHz
3s
4
6
8.2.2016
Average Spectrum Occupancy –
Selected Bands across 5 Sites
Reference Architecture: WiFiUS Observatories / Data store
IIT Chicago / VT
Analysis
Server
Capture
RFEye
Capture
PC
RFEye
Capture
PC
RFEye
RF Measurement System
PC
U.S.
RF Measurement System
RF Measurement System
Turku
Finland
Long-Term
Measurement
Storage Server
PC
- Local Graph
and Plots
Web
Server
Internet
Oulu / Turku
Analysis
Server
Capture
RFEye
Capture
PC
RFEye
PC
RF Measurement System
IIT
IITSO Long
Term
Measurements
Long-Term
Measurement
Storage Server
RF Measurement System
RFEye
Capture PC
Local Analysis
Analysis Server
Storage Server
OS
Linux
Windows
(Linux)
Windows
Windows
Windows
Apps
• Logger
• NSF Store
on PC
• Log file
store
• Upload to long
term
• Browser
• RFeye Live
• RFeye View
• Roberson
Analysis SW
• APP.PY
Web Server
Measurement
Format
•Native
RFeye
• Native RFeye
• Status Log files
•Long term format
Size in
Memory
•Read long term
files
SW Library
DBase
(RFeye
Binary Files)
(RFeye Binary
Files)
Mongo / dsNet
(16 TBytes / 96 TB)
7
8.2.2016
WiFiUS Server System
• Standard rack system
• Cleversafe dsNet storage
• High capacity + robustness
• Dedicated analysis
machine
• Specifications
• 2.0 GHz - 12 core
Xeon Processor
• 128 GB RAM
• ~200 TB storage
• 10 Gbit Internet
connection
Non-Stationary Hidden Markov Models
• VTT developed applications of NonStationary Hidden Markov Models to
predicting spectrum occupancy
(Chen, et Al., IEEE Wireless
Communication Letters, 2014).
• Current results show flexibility and
good predictive performance.
• Ongoing work at VT to:
• Reduce computational complexity of
model,
• test limits of model applicability (in
different bands and applications), and
• apply similar models to cellular
network load prediction.
8
8.2.2016
Data Fusion
•
Spectrum data
Study of Chicago long-term spectrum data (2009-) concentrating on both
large-scale (whole band sweep) and more specific bands (e.g., land
mobile radio and cellular bands). This has now been expanded and
complemented by the global RFeye data produced through
measurements at various sites in the US and Finland.
•
Open data
Weather data (http://www.wunderground.com/): extracting
temperature, wind speed, snow depth, precipitation etc. from the
weather stations near the spectrum observatories
Chicago data portal (https://data.cityofchicago.org/): Chicago park
district events, public libraries Wi-Fi usage, & other ”smart city” datasets
Mass event data: extracting mass event information (e.g. baseball,
football, soccer, and athletics) near spectrum observatories and
specifically on-site during the game (e.g., baseball)
•
Data fusion of multiple sources
Pre-processing, aggregating, filtering, and synchronization of RF
spectrum data, open data and game data
Large-scale statistical analysis and modelling
Analysis and modelling
• Bayesian non-parametric models based on Gaussian process (GP)
priors
• Time series modeling of usage patterns: short non-periodic and longterm trends, daily, weekly, yearly seasonal effects of spectrum, weather,
and other related open data
• Hierarchical Bayesian linear regression analysis of spectrum (using
game data)
• Regression analysis of different variables correlated with spectrum
• We can efficiently model multi-level structure of data and its uncertainty:
different sites, frequency bands, time resolution etc. simultaneously
• Implementation using R and Stan probabilistic programming language
• Analysis of mass event / game data: Can we infer changes of cellular
bands during mass events: spectrum measurements as response
variable and game timestamps, event size, event distance from
observatories, frequency band label, site label, time of day, day of week
as predictors.
9
8.2.2016
Infrastructure Efforts
Big Data
• Implementation of parallel distributed computing environment based
on R, Hadoop, and MapReduce for effective processing of largescale datasets (such as spectrum, mass events, and open datasets)
• e.g. - Master thesis: Jaakko Lampi (2014) "Large-Scale Distributed
Data Management and Processing Using R, Hadoop and
MapReduce", Dept. of Computer Science and Engineering,
University of Oulu
Guidelines developed for spectrum occupancy
measurements
• Comprehensive survey of spectrum occupancy measurements
including key concepts, objectives, phases, measurement systems,
guidelines and research challenges.
Significant Contributions & Achievements
• Academic Achievements
IIT – 2 PhDs (Tanim Taher) / 3 M.S.
University of Oulu – 1 M.S.
Virginia Tech – 1 M.S. (1 PhD - 2016)
• Critical Government Contributions
Chaired + Participation at WSRD meeting in US on spectrum
measurements and presentation of U.S. and European activities.
Spectrum occupancy metrics were contributed to ITU-R WP5A
work on cognitive radio systems (CRS) in the land mobile service in
collaboration with industry and Finnish CORE+ project including
chairmanship of CRS sub-working group.
On-going engagement with both the FCC and NTIA
• Student exchange and visits
Turku graduate student at IIT for 3 months / IIT graduate student
at Turku / Oulu – 6 weeks
• Met every two weeks for nearly three years!
10
8.2.2016
WiFiUS Related Research Based Papers
Marja Matinmikko, Miia Mustonen, Dennis Roberson, Jarkko Paavola, Marko
Höyhtyä, Seppo Yrjölä, and Juha Röning. Overview and comparison of recent
spectrum sharing approaches in regulation and research: From opportunistic
unlicensed access towards licensed shared access, IEEE DySPAN 2014,
Washington, D.C.
Marko Höyhtyä, Marja Matinmikko, Xianfu Chen, Juhani Hallio, Jani Auranen, Reijo
Ekman, Juha Röning, Jan Engelberg, Juha Kalliovaara, Tanim Taher, Ali Riaz, and
Dennis Roberson. Measurements and Analysis of Spectrum Occupancy in the 2.32.4 GHz band in Finland and Chicago, CrownCom 2014, Oulu, Finland
Tanim Taher, Ryan Attard, Ali Riaz, Dennis Roberson, Jesse Taylor, Kenneth J
Zdunek, Juhani Allio, Reijo Ekman, Jarkko Paavola, Jaakko Suutala, Juha Roning,
Marja Matinmikko, Marko Hoyhtya, Allen B. MacKenzie, Global Spectrum
Observatory Network Setup and Initial Findings, CrownCom 2014, Oulu, Finland
Ryan Attard, Juha Kalliovaara, Tanim Taher, Jesse Taylor, Dennis Roberson, A
High-performance Tiered Storage System for a Global Spectrum Observatory
Network, CrownCom 2014 – WiFiUS Workshop, Oulu, Finland
Miia Mustonen, Marja Matinmikko, Dennis Roberson, Seppo Yrjola, Evaluation of
recent spectrum sharing models from the regulartory point of view, First
International Conference on 5G for Ubiquitous Connectivity, 2014 Levi, Finland
WiFiUS Related Research Based Articles
Xianfu Chen, Honggang Zhang, Allen B. MacKenzie, and Marja Matinmikko.
Predicting Spectrum Occupancies Using a Non-Stationary Hidden Markov Model,
IEEE Wireless Communications Letters, 2014
Marko Hoyhtya, Marja Matinmikko, Xianfu Chen, Juhani Hallio, Jani Auranen, Reijo
Ekman, Juha Roning, Jan Engelberg, Juha Kalliovaara, Tanim Taher, Ali Riaz,
Dennis Roberson, Spectrum Occupancy Measurements in the 2.3-2.4 GHz band:
Guidelines for Licensed Shared Access in Finland, Cognitive Communications, 28
May 2015
Abdallah S. Abdallah, Allen B. MacKenzie, Vuk Marojevic, Roger B Bacchus, Ali
Riaz, Dennis Roberson, Juha Kalliovaara, Juhani Hallio, Reijo Ekman, Detecting
the Impact of Human Mega-Events on Spectrum Usage, IEEE CCNC 2016, Las
Vegas, USA, January, 2016, (accepted)
Marko Höyhtyä, Aarne Mämmelä, Marina Eskola, Marja Matinmikko, Juha
Kalliovaara, Jaakko Ojaniemi, Jaakko Suutala, Reijo Ekman, Roger Bacchus,
Dennis Roberson, Spectrum occupancy measurements: A survey and use of
interference maps, IEEE Communications Surveys & Tutorials (COMST Submitted)
Tanim Taher, Dennis Roberson, SYSTEM AND METHOD FOR DETERMING
AND SHARING SPECTURM AVAILABILITY - IIT-293-P, Filed Patent, Sept 2014
11
8.2.2016
Application: Spectrum Assignment Roadmap and
Dynamic Spectrum Access
Dynamic Spectrum Access
Selection of operating frequency,
protocols, and transmission parameters
dynamically in real-time based on RF
environment and policy.
TODAY
Pre-arranged,
Opportunistic
Spectrum access
Based on primary
User characteristics
Pre-arranged, dynamic
Spectrum assignment
based on geo-location
database
Static Frequency
Assignment
Unrestricted
Opportunistic
Spectrum access
And Real-Time
Spectrum Negotiation
time
Dynamic Spectrum Sharing
Dynamic Frequency Selection
Approaches
Listen-before-talk
TV Whitespace
Radar Band Sharing
Examples
450-470 MHz LMR Band – Chicago 2012
12
8.2.2016
Architecture View for WiFiUS Observation System
Architecture View for WiFiUS Observation System
13