Outline Introduction M ulti-Channel Reliability and Spectrum - PDF document

Outline  Introduction  M ulti-Channel Reliability and Spectrum Usage Experimental methodology  Empirical study in homes in Real Homes  Spectrum study of existing wireless signals Empirical Studies for Home-Area Sensor Networks  802.15.4 link reliability in all 16 channels  Conclusion M o S ha, Gregory Hackmann, Chenyang Lu Department of C omputer Science and Engineering 2 Smart Grid Wireless Sensor Networks   Home Area Network Advantage  Power meters, smart thermostats, home appliances.  Do not require wired infrastructure.   Easily and inexpensively retrofit existing homes. Enables both wired and wireless communication between  Energy efficiency utility companies and household devices  Reliability challenges  Crowded 2.4 GHz IS M band  Unpredictable environment 3 4 Outline M ethodology   Introduction Spectrum usage between 2.400 GHz and 2.495 GHz   Wi-Spy 2.4x spectrum analyzer Experimental methodology   Sweep across the 2.4 GHz spectrum Empirical study in homes  Sampling period: 40 ms  Spectrum study of existing wireless signals  Signal strength reading on each of 254 discrete frequencies  802.15.4 link reliability in all 16 channels  Traces over 7 days in 6 apartments and Bryan Hall  Conclusion  Normal daily activities  15,120,000 readings for each of the 254 frequencies  2.5 GB of data per location  Convert signal strength readings to binary values based on a threshold  Communication theory:  0: idle channel  1: busy channel 5 6

Spectrum Usage Traces Questions  Collected from the 2.4 GHz spectrum in six apartments and an office 1. Is there a channel free in all apartments? No. There is no “golden” channel . 2. Do homes have similar spectrum usage patterns as offices? No. Test in lab is not enough . 3. Does spectrum usage change over time? Yes. Channel configuration won’t work. 4. Is 802.11 the dominant interferer in homes? 7 8 Is Wi-Fi the dominant user of the spectrum? Is Wi-Fi the dominant user of the spectrum? 9 10 Is Wi-Fi the dominant user of the spectrum? Is Wi-Fi the dominant user of the spectrum?   While Wi-Fi is a major source of interference, others can While Wi-Fi is a major source of interference, others can be non-negligible contributors to spectrum occupancy. be non-negligible contributors to spectrum occupancy. Channels overlapping with Wi-Fi channel Channels overlapping with Wi-Fi channel Channels not overlapping with Wi-Fi channel Channels not overlapping with Wi-Fi channel 11 12

Outline M ethodology   Platform Introduction   Tmote Sky and TelosB motes Experimental methodology  IEEE 802.15.4 compliant Chipcon CC2420 radio  Empirical study in homes  16 channels (11—26) in 5 MHz steps  Spectrum study of existing wireless signals  TinyOS2.1 using default CSMA/ CA M AC layer  802.15.4 link reliability in all 16 channels  Conclusion  Packet Reception Rate (PRR) of all 802.15.4 channels  10 apartments, 24 hours per apartment  A node broadcast 100 packets per channel to multiple receivers, cycling through all 16 channels in 5 minutes  Receivers recorded the PRRs in onboard Flash 13 14 Questions Is there a persistently reliable channel?  1. Is there a persistently reliable channel? Link reliability varies among apartments and links. 2. If a good channel cannot be found, are retransmissions sufficient to deal with packet loss? 3. If no single channel can be used for reliable operation, can we exploit channel diversity to achieve reliability? 4. Do channel conditions exhibit cyclic behavior over time? 5. Is reliability strongly correlated among different channels? Different apartments Different links within a same apartment 15 16 Is there a persistently reliable channel? Is there a persistently reliable channel?   Link reliability varies among apartments and links. Link reliability varies among apartments and links. Different apartments Different apartments Different links within a same apartment Different links within a same apartment 17 18

Is retransmission sufficient? Is retransmission sufficient?   No, due to burstiness of transmission failures. No, due to burstiness of transmission failures. 10 % of time, consecutive drops larger than 60 19 20 Is channel hopping effective? How often needs a link switch channel?   Yes! Only a small number of channel hops per day. Number of channel hops required under an optimal schedule Optimal channel hopping schedule (one link selected randomly per apartment) Single best channel 21 22 Findings Questions  1. Is there a persistently reliable channel? Home environments are much more complicated than offices  numerous and diverse Wi-Fi APs and others. 2. If a good channel cannot be found, are retransmissions  sufficient to deal with packet loss? There is no channel that works for all homes  no channel works for all. 3. If no single channel can be used for reliable operation, can  we exploit channel diversity to achieve reliability? Channel reliability changes dynamically 4. Do channel conditions exhibit cyclic behavior over time?  cannot pre-select channels.  Channel hopping is effective No. The channel behavior is not cyclic. 5. Is reliability strongly correlated among different channels?  enhance reliability with a few channel switches per day. Yes. Avoid using adjacent channel. 23

Outline  Introduction  Related work ARCH: Practical Channel Hopping for Reliable  Protocol design Home-Area Sensor Networks  Design Insights  ARCH Protocol Outline  Coordinated Channel Hopping  Handling Channel Desynchronization M o S ha, Gregory Hackmann, Chenyang Lu  Evaluation Department of C omputer Science and Engineering  Conclusion 26 ARCH Protocol Outline Coordinated Channel Hopping  Receiver-oriented protocol  Upon selecting a new channel, nodes notify their neighbors  M onitor channel condition of this change. Neighbors update their neighbor tables.  M aintain a sliding window of ETX values of incoming links  M ulti-hop problem  M ark channel unreliable if ETX values exceed threshold  Blacklist current channel when channel condition degraded  Switch to a new picked channel  There is strong correlation among adjacent channels  M ulti-sender problem  Uses a probabilistic scheme  Generate a random number  If q falls into the range then, switch to channel i 27 28 Handling Channel Desynchronization Outline   Default channel Introduction   No data transmission, only for resynchronization Related work  Senders use when reach maximum retransmissions  Protocol design  Receivers use when reach maximum waiting time  Evaluation  False detection  Simulator-Based M icrobenchmarks  Channel is too noisy * Channel Selection Scheme  Exchange previous channels when resynchronizing * Channel Quality Estimator  Blacklist this channel and pick a new one  Real-World M acrobenchmarks * Single-Hop Data Collection * M ulti-Hop Data Collection  Conclusion 29 30

Simulator-Based M icrobenchmarks Simulator-Based M icrobenchmarks   Channel Selection Scheme Channel Selection Scheme 31 32 Simulator-Based M icrobenchmarks Simulator-Based M icrobenchmarks   Channel Quality Estimator Channel Quality Estimator 33 34 Real-World M acrobenchmarks Real-World M acrobenchmarks   Single-Hop Data Collection Single-Hop Data Collection 35 36

Real-World M acrobenchmarks Real-World M acrobenchmarks   M ulti-Hop Data Collection M ulti-Hop Data Collection 37 38 Real-World M acrobenchmarks Real-World M acrobenchmarks   M ulti-Hop Data Collection M ulti-Hop Data Collection 39 40 Conclusion  ARCH has the following salient features that distinguish it from existing channel diversity schemes:  Adaptively channel selection  Distributed approach  Lightweight and robust  M inimal communication overhead  Results  42.3% decrease in packet retransmissions  17% increase in the proportion of links with perfect delivery rates  2.2X increasing in the minimum delivery rate for the most challenging of links  31.6% average reduction in radio usage