WiOpt2015_v1.0
Download as PPTX, PDF0 likes515 views
1) The document discusses using unused UHF TV spectrum in India for providing rural broadband connectivity. 2) It analyzes that about 100MHz of spectrum in the 470-590MHz band is unused. 3) An optimization tool is proposed that determines optimal transmit power levels, link activity schedules, and routing to maximize throughput for a rural village broadband network using this spectrum.
![UHF Band-IV (470-590MHz) Utilization in India
Our Analysis of available spectrum
3 September 28, 2015
Band Characteristics
2
15 channels of 8MHz
each
3
At any place at least 12
out of 15 channels are
always available
1
Primary user: Doordarshan
373 transmitters overall
4
Better propagation
characteristics than
existing unlicensed
band
5
Potential for providing
affordable rural
broadband
* Using protection/pollution viewpoint [Mishra-
Sahai’09]
Our analysis reveals about 100MHz unused in
UHF Band-IV](https://image.slidesharecdn.com/7cc4f9ce-e842-473b-913d-58db4fa3e86e-150928032318-lva1-app6892/85/WiOpt2015_v1-0-3-320.jpg)
![Simulation Results
Transmit Power requirement:
18 September 28, 2015
*[Bandwidth required for ISPs for better connectivity and improved quality of service,
Tech. Rep. 2009, TRAI]
Total population: 2524
No of people in village: 1200
No of people in hamlets:1324
17 dBm transmit power
i.e., 27 dBm EIRP
is sufficient for providing
broadband services to all
users at hamlets
*(contention ration 50)
No. of served users = No. of
active users * contention
ratio](https://image.slidesharecdn.com/7cc4f9ce-e842-473b-913d-58db4fa3e86e-150928032318-lva1-app6892/85/WiOpt2015_v1-0-18-320.jpg)
![Performance Evaluation
Assumptions and Model:
24 September 28, 2015
• Assumptions:
– Timeframes consists of slots of fixed duration
– Packet length depends on link capacity and timeslot duration
• Model:
– Buffer content at 𝑖 𝑡ℎ node during 𝑛, 𝑛 + 1 timeslot interval:
𝑏𝑖[𝑛 + 1] =
𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 − 𝑢 𝑏𝑖 𝑛 𝑖𝑓 𝑛𝑜𝑑𝑒 𝑖 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑠;
𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 + 𝑢 𝑏𝑖 𝑛 𝑖𝑓 𝑛𝑜𝑑𝑒 𝑗 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑠
𝑎𝑛𝑑 𝑛𝑜𝑑𝑒 𝑖 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑠;
𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒.
* 𝑏𝑖 𝑛 : Content of buffer at 𝑖 𝑡ℎ
in 𝑛 𝑡ℎ
timeslot
* 𝑦𝑖 𝑛 : Number of packets entering in 𝑖 𝑡ℎ
node’s buffer during (n-1,n] timeslot interval.
*Function u(l) is defined as u(l)=1,l>0 and 0 otherwise
Back](https://image.slidesharecdn.com/7cc4f9ce-e842-473b-913d-58db4fa3e86e-150928032318-lva1-app6892/85/WiOpt2015_v1-0-24-320.jpg)
![Hill Climbing Algorithm
Solving non-convex optimization
26 September 28, 2015
• Search for power vector 𝑃 0 such that optimization
constraints satisfied
• Considering a positive number sequence 𝜂 𝑘
– Perturb 𝑃 0 in 𝑃𝑖 0 − 𝜂 𝑘, 𝑃𝑖 0 + 𝜂 𝑘 for all nodes during
iteration 𝑘.
– Compute vector 𝑃𝑡𝑒𝑚𝑝
– We get the next vector in sequence 𝑃 1 = 𝑃𝑡𝑒𝑚𝑝 if
𝑖=1
𝑁
𝑎𝑖,𝑡𝑒𝑚𝑝 𝑃𝑡𝑒𝑚𝑝,𝑖 < 𝑖=1
𝑁
𝑎𝑖 0 𝑃𝑖[0]
• Subsequent vectors 𝑃 1 , 𝑃 2 and so on obtain power
vectors closer to optimal
• 𝜂 𝑘 is reduced as 𝑘 increases.](https://image.slidesharecdn.com/7cc4f9ce-e842-473b-913d-58db4fa3e86e-150928032318-lva1-app6892/85/WiOpt2015_v1-0-26-320.jpg)
WiOpt2015_v1.0
- 1. How Much TV UHF Band Spectrum is Sufficient for Rural Broadband Coverage? September 28, 2015 Abhay Karandikar Department of Electrical Engineering Indian Institute of Technology Bombay, Mumbai 400076 Joint work with: Animesh Kumar, Rajeev Kumar, Punit Rathod, Abhay Karandikar
- 2. Terrestrial TV Spectrum Allocations Global usage and the Indian scenario 2 September 28, 2015 862-890 Fixed, Mobile, Except Aeronautical Mobile, Broadcasting 470-790 Broadcasting 790-862 Fixed, Broadcasting, Mobile Except Aeronautical Mobile 608-614 Radio Astronomy, Mobile- satellite Except Aeronautical Mobile-satellite (Earth-to- space) 470-512 Broadcasting, Fixed, Mobile 512-608 Broadcasting 610-890 Fixed, Mobile, Broadcasting 470-585 Fixed, Mobile, Broadcasting 585-610 Fixed, Mobile, Broadcasting, Radio navigation 614-698 Broadcasting, Fixed, Mobile 698-806 Broadcasting, Fixed, Mobile 806-890 Fixed Mobile Broadcasting Region 1 (Europe, Africa, Russia, Middle East) Region 2 (Americas, Pacific) Region 3 (India - Asia, Oceania) • Government’s national broadcaster named Doordarshan holds all of the terrestrial TV broadcasting license • ITU Regulations for Region 3 (applies to India) allows use of 470-585 MHz for “Fixed, Mobile, and Broadcasting” as Primary Services • FCC (US), Ofcom (UK), have permitted the use “white spaces” on secondary basis
- 3. UHF Band-IV (470-590MHz) Utilization in India Our Analysis of available spectrum 3 September 28, 2015 Band Characteristics 2 15 channels of 8MHz each 3 At any place at least 12 out of 15 channels are always available 1 Primary user: Doordarshan 373 transmitters overall 4 Better propagation characteristics than existing unlicensed band 5 Potential for providing affordable rural broadband * Using protection/pollution viewpoint [Mishra- Sahai’09] Our analysis reveals about 100MHz unused in UHF Band-IV
- 4. Current Status of Broadband in Rural India 4 September 28, 2015 • Government of India announced National Optical Fiber Network (NOFN) to link all Gram Panchayats with optical connectivity. • Leveraging on the NOFN, – UHF Band-IV to provide affordable broadband in (rural) India • Summary statistics of NOFN / Gram Panchayats Number of Blocks (NOFN Phase-I) 6,382 Number of Gram Panchayats (NOFN Phase I/II) 2,50,000 Number of Villages 6,38,619 Avg. number of Gram Panchayats per block 40 Avg. number of Villages per Gram Panchayat 2.56 Avg. number of Hamlets per Village 4
- 5. Rural Broadband Penetration 5 September 28, 2015 District, City, Block Existing Fiber Connectivity Gram Panchayat (GP) Gram Panchayat (GP) Fiber Connectivity By NOFN Village ( Distance from GP 2-6 Km ) Village Village Village Still, No Internet Access !!!
- 6. TV Band Multihop Mesh Network Point of Presence (PoP) with NOFN connectivity 6 September 28, 2015 Village Village Village Hamlet Hamlet Hamlet Hamlet Hamlet Fiber Optic PoP
- 7. Throughput Optimization Tool 7 September 28, 2015 Optimization Tool Sink Nodes (PoP) Rural Clusters (APs) Demography Information Available Frequencies Path Loss Models Optimal power And routes Objective of the Optimization: Determine optimal power, link activity time and routes We consider both Generalized routing and destination based routing
- 8. Optimization Tool Input Offered Load 8 September 28, 2015 Village Village Fiber Optic PoP Data generated at each node = 𝛼𝑅𝑁𝑖 Fraction of users Served at each village Population of Village / Hamlet Data Rate from each user Load at each node = 𝛼𝑅𝑁𝑖 + Data generated at neighors
- 9. Interference Zone 9 September 28, 2015 1 3 6 5 2 4 7 8 Interference Zone for nodes 2,4,5,7 Links • 7 to 8 • 5 to 4 • 2 to 3 Can operate simultaneously Link 4 to 8 Interferes with all other links Node 1 sends 𝛼𝑁1 𝑅 bits/secNode 3 sends 𝛼(𝑁1+𝑁2 + 𝑁3)𝑅 bits/sec
- 10. Destination Based Routing Optimization Setup Generalized routing and Destination based routing 10 September 28, 2015 Minimize 𝑖∈Τ 𝑎𝑖 𝑃𝑖 Subject to 𝑥∈𝑋 𝑖 𝜆 𝑥,𝑖 𝐶 𝑥𝑖(𝑃𝑥) + 𝛼𝑅𝑁𝑖 = 𝑦∈𝑌 𝑖 𝜆𝑖,𝑦 𝐶𝑖𝑦(𝑃𝑖) for all 𝑖 ∈ 𝑇 𝜆 𝑥,𝑖 ≥ 0 for all 𝑖 ∈ 𝑇, 𝑥 ∈ 𝑋𝑖 𝜆𝑖,𝑦 ≥ 0 for all 𝑖 ∈ 𝑇, y ∈ 𝑌𝑖 𝑢∥𝑣∈𝐼𝑍 𝑖,𝑗 𝜆 𝑢,𝑣 ≤ 1 for all 𝑖 ∈ 𝑇, j ∈ 𝑌𝑖 𝑃𝑖,𝑚𝑖𝑛 ≤ 𝑃𝑖 ≤ 𝑚𝐸𝐼𝑅𝑃 Power Consumed: Active time fraction and Power for all nodes Flow Constraint: Incoming traffic + Traffic generated at node = Outgoing traffic Interference avoidance using TDMA: Fraction of time active within each interference zone to be less than 1 Maximum Power: Maximum power for each link 𝑥∈𝑋 𝑖 𝜆 𝑥,𝑖 𝑑 𝑥𝑖 𝐶 𝑥𝑖(𝑃𝑥) + 𝛼𝑅𝑁𝑖 = 𝑦∈𝑌 𝑖 𝜆𝑖,𝑦 𝑑𝑖𝑦 𝐶𝑖𝑦(𝑃𝑖) for all 𝑖 ∈ 𝑇 Generalized Routing STATE OF INCOMING LINKS: 𝑑 𝑥𝑖 = 1 if node x routes packets via node 𝑖; 0 Otherwise. STATE OF OUTGOING LINKS: 𝑑𝑖𝑦 ∈ 0,1 and, 𝑑𝑖𝑦 = 1.
- 11. Optimization Steps Solving non-convex optimization problem 11 September 28, 2015 Assign Initial transmit power Determine link connectivity Pathloss Model Interference computation Initial power vector 𝑃 0 Perturb 𝑃 𝑘 − 1 in cube 𝑃𝑖 𝑘 − 1 − 𝜂 𝑘, 𝑃𝑖 𝑘 − 1 + 𝜂 𝑘 to get 𝑃𝑡𝑒𝑚𝑝 Solve using Linear Programming to obtain activity times 𝑎𝑖,𝑡𝑒𝑚𝑝 𝑃 𝑘 = 𝑃𝑡𝑒𝑚𝑝 if solution better Hill Climbing Reduce 𝜂 𝑘 as 𝑘 reaches threshold 𝑘 = 𝑘 + 1 𝑘 = 𝑘 + 1
- 12. Example Topology Location of the nodes in Dhuktan 12 September 28, 2015 Area: 6.25 sq. km. Population: 2524 Dhuktan: 1200 Other Hamlets: 1324
- 13. Testbed Evaluation Parameters 13 September 28, 2015 Parameters Description Value 𝑓𝑐 Carrier frequency 500 MHz 𝑊 *Bandwidth of the channel 8 MHz 𝑅 Target per user broadband speed 2 Mbps ℎ𝑖 Sink node antenna height 40 m ℎ𝑗 Transmission node antenna height 40 m 𝐺 𝑇𝑥 Transmitter antenna gain 10 dBi 𝜉𝑖 Receiver sensitivity of transmission node radios -86 dBm 𝑁0 Thermal noise for 8MHz bandwidth -104.75 dBm 𝛻 SNR threshold to determine whether a node is in the interference zone -5 dB *8 MHz bandwidth of the channel is considered for uplink and downlink separately.
- 14. Routing Results Generalized Routing vs Destination based Routing: 14 September 28, 2015 Low network load: Generalized routing Hamlet name (number of users in the hamlet, transmit power, distance from sink node) Throughput of the link in Mbps Algorithm distributes load at every node *Above routing result considers 14/1324 active users simultaneously.
- 15. Routing Results Generalized Routing vs Destination based Routing: 15 September 28, 2015 Low network load: Destination based routing Algorithm selects best path from each node to sink *Above routing result considers 14/1324 active users simultaneously.
- 16. Routing Results Generalized Routing vs Destination based Routing: 16 September 28, 2015 High network load: Generalized routing Algorithm selects best path from each node to sink *Above routing result considers 40/1324 active users simultaneously.
- 17. Routing Results Generalized Routing vs Destination based Routing: 17 September 28, 2015 High network load: Destination based routing Algorithm selects best path from each node to sink High network load: both Generalized routing and Destination based routing selects same routing table *Above routing result considers 40/1324 active users simultaneously.
- 18. Simulation Results Transmit Power requirement: 18 September 28, 2015 *[Bandwidth required for ISPs for better connectivity and improved quality of service, Tech. Rep. 2009, TRAI] Total population: 2524 No of people in village: 1200 No of people in hamlets:1324 17 dBm transmit power i.e., 27 dBm EIRP is sufficient for providing broadband services to all users at hamlets *(contention ration 50) No. of served users = No. of active users * contention ratio
- 19. Throughput Expected throughput vs Achievable throughput 19 September 28, 2015 Achievable throughput depends upon: Load in the network Number of hops Low throughput at low load: Number of hops high Packet length determined by link having lowest capacity in the path High packet loss at forwarding nodes Throughput vs Percentage of served users Computation Details
- 20. Summary 20 September 28, 2015 • A middle-mile multi-hop mesh network in UHF TV Band evaluated for providing connectivity to rural underserved areas • Optimization tool utilizes demographic information to provide optimal – Power allocation – Link activity times – Routing of traffic • Modular setup for the optimization tool that allows different pathloss and topology descriptions • For a village Dhuktan in India, 17dBm transmit power with 10dBi antenna sufficient to provide 2Mbps connectivity to entire population of 1324 with a contention ratio of 50.
- 21. Proposed Test-bed at Palghar in Maharashtra 21 September 28, 2015 30 m 10 m UHF Link (2.5 Km) WiFi Hotspot UHF Link (0.7 – 1.2 Km) WiFi Coverage (~ 100 – 200 m) 2-3 m Dhuktan Pada Khamloli Village Bahadoli Pada 700 m 500 m 2.3 Km 600 m 550 m Bahadoli Village P2P Wi-Fi link Khamloli Tower
- 22. Thank You
- 23. Backup Slides
- 24. Performance Evaluation Assumptions and Model: 24 September 28, 2015 • Assumptions: – Timeframes consists of slots of fixed duration – Packet length depends on link capacity and timeslot duration • Model: – Buffer content at 𝑖 𝑡ℎ node during 𝑛, 𝑛 + 1 timeslot interval: 𝑏𝑖[𝑛 + 1] = 𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 − 𝑢 𝑏𝑖 𝑛 𝑖𝑓 𝑛𝑜𝑑𝑒 𝑖 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑠; 𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 + 𝑢 𝑏𝑖 𝑛 𝑖𝑓 𝑛𝑜𝑑𝑒 𝑗 𝑡𝑟𝑎𝑛𝑠𝑚𝑖𝑡𝑠 𝑎𝑛𝑑 𝑛𝑜𝑑𝑒 𝑖 𝑟𝑒𝑐𝑒𝑖𝑣𝑒𝑠; 𝑏𝑖 𝑛 + 𝑦𝑖 𝑛 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒. * 𝑏𝑖 𝑛 : Content of buffer at 𝑖 𝑡ℎ in 𝑛 𝑡ℎ timeslot * 𝑦𝑖 𝑛 : Number of packets entering in 𝑖 𝑡ℎ node’s buffer during (n-1,n] timeslot interval. *Function u(l) is defined as u(l)=1,l>0 and 0 otherwise Back
- 25. Optimization Steps 25 September 28, 2015 • Step 1: – Assign initial transmit power – Pathloss model and receive power criteria determine link establishment – Form connectivity graph and determine interference • Step 2: – Interference avoidance using TDMA and spatial reuse
- 26. Hill Climbing Algorithm Solving non-convex optimization 26 September 28, 2015 • Search for power vector 𝑃 0 such that optimization constraints satisfied • Considering a positive number sequence 𝜂 𝑘 – Perturb 𝑃 0 in 𝑃𝑖 0 − 𝜂 𝑘, 𝑃𝑖 0 + 𝜂 𝑘 for all nodes during iteration 𝑘. – Compute vector 𝑃𝑡𝑒𝑚𝑝 – We get the next vector in sequence 𝑃 1 = 𝑃𝑡𝑒𝑚𝑝 if 𝑖=1 𝑁 𝑎𝑖,𝑡𝑒𝑚𝑝 𝑃𝑡𝑒𝑚𝑝,𝑖 < 𝑖=1 𝑁 𝑎𝑖 0 𝑃𝑖[0] • Subsequent vectors 𝑃 1 , 𝑃 2 and so on obtain power vectors closer to optimal • 𝜂 𝑘 is reduced as 𝑘 increases.
- 27. TV Band Utilization in India and the Digital Divide 27 September 28, 2015 • One National broadcaster (Doordarshan) with at-most 2 channels • In 470-590 MHz UHF TV Band, least 56.27% areas in India, all the 15 channels (100% of the TV UHF band) are free • Government of India’s project National Optical Fiber Network (NOFN) to link all gram panchayats Number of Blocks 6,382 Number of Gram Panchayats 2,50,000 Number of Villages 6,38,619
- 28. Delay and Jitter 28 September 28, 2015 Result similar to results of Achievable throughput result Same explanation can be followed Delay and Jitter vs percentage of served users Error Bar shows the confidence interval of 95%