Enabling Reliable, Asynchronous, and Bidirectional Communication in SNOW Abusayeed Saifullah*, Mahbubur Rahman*, Dali Ismail, Chenyang Lu, Jie Liu, Ranveer Chandra *co-primary author Wayne State University Washington University in St Louis Microsoft Research

Low-Power Wide-Area Network (LPWAN) Overcomes scalability and range limits of traditional wireless sensor networks. A key technology driving the Internet of Things (IoT). LoRa SigFox nWave RPMA NB-IoT EC-GSM-IoT LTE Cat M1 5G

SNOW (Sensor Net. over White Space) LPWAN by exploiting the TV white spaces. White Space: unused TV channels between 54-698MHz. SNOW 1.0 [SenSys ’16, Best Paper Nominee] Long range  nodes directly Tx to the base station (BS). The BS accesses cloud for white space database. Spectrum split into narrow orthogonal subcarriers. Each node transmits on a subcarrier. Multi-Rx at the BS using a single antenna-radio.

Decoding for Multi-Rx at BS Aggregate OFDM signal in time domain Orthogonal signals from sensor nodes on orthogonal subcarriers.

SNOW 2.0 Uplink: Multi-Rx with single-antenna radio (SNOW 1.0). Downlink: Multi-Tx with single-antenna radio Send different data to different nodes using a single Tx Bidirectional: concurrent uplink and downlink comm. Fully Asynchronous Reliability ACK for each Tx

SNOW 2.0 Dual-Radio Design Simultaneous Tx/Rx at BS is enabled by using two radios. One radio is dedicated for Tx, the other for Rx. BS uses wide white space spectrum Each node uses one half-duplex narrowband radio

PHY Design Using D-OFDM D-OFDM: Distributed OFDM Tx on narrow OFDM subcarriers  energy and spectrum efficiency Individual subcarrier modulation: BPSK Uplink Each node independently encodes and transmits. BS runs decoder based on global FFT to decode asynchronous Tx. Downlink: needs an inverse implementation. Different from MIMO radio that rely on multiple antennas Single antenna is important for low frequency band (54-698MHz).

Downlink Communication BS encodes different data on different subcarriers. Different from traditional broadcast. Performs IFFT. Makes a single transmission. From the received OFDM signal, a node independently decodes data from signal component on its subcarrier. Can encode anytime on a subcarrier Distributed decoding

MAC Protocol White space spectrum split into orthogonal subcarriers. Each node is assigned a subcarrier. When number of nodes> subcarriers, a subcarrier is shared. Location-aware subcarrier allocation Attempt to assign different subcarriers to hidden terminals. Also ensure a subcarrier is not overly congested. Tx using a lightweight CSMA/CA protocol (like TinyOS) Static interval random backoff Try to assign different subcarriers to u and w. u w

MAC Protocol: Handling ACK D-OFDM allows us to encode data anytime on any subcarrier  allows ACK of asynchronous Tx. Rx radio keeps receiving while Tx radio keeps sending. When there is nothing to transmit, Tx radio can sleep. CSMA/CA helps avoid interference on ACK. ACK for u will be sensed by v. u v

Other Features Peer to peer communication Handling dynamics Spectrum dynamics through backup subcarrier Subcarrier swap Load balancing Node join/leave SNOW can exploit fragmented white space spectrum. Fragmented white space spectrum Signal strength (dBm) Frequency (MHz)

Implementation Implemented in GNU Radio Experiment with USRP device Connected to laptop or Raspberry pi All the packets (up to # of subcarriers) are decoded within 0.1ms  comparable to a single packet decoding time scalability.

Design Parameters Key design parameters Default settings Bit rate: target bit rate at least 50kbps. Packet size Subcarrier bandwidth Bit spreading factor Default settings Tx power: 0dBm Subcarrier bandwidth: 400kHz BS bandwidth: 6MHz Packet size: 40bytes Bit spreading factor: 8 Determined based on target bit rate, Shannon-Hartley Theorem, Nyquist Theorem, and experiment.

Urban Deployment Deployment in Detroit, Michigan

Distance vs. Tx Power 400kHz bandwidth Approx. 8km at 0dBm

Reliability Uplink Downlink 400kHz 0 dBm bandwidth SNOW is reliable in both directions

Max Achievable Throughput Throughput increases linearly with the number of subcarriers due to parallel receptions at BS.

Indoor Deployment Lower frequency propagates well through obstacles. Deployment in CS building in Wayne State University

Rural Deployment (Rolla, Missouri) Throughput increases linearly with the number of subcarriers due to parallel receptions at BS.

Benefits of SNOW over Other LPWANs Limited infrastructure NB-IoT, 5G, LTE Cat M1, EC-GSM-IoT need infrastructure. Bidirectional Most non-cellular LPWANs are primarily uplink only. Scalability increases with the spectrum availability due to high parallelism. LoRa, SigFox achieve scalability assuming very low traffic. Abundant spectrum Spectrum availability in the US counties

Comparison with LoRa (QualNet Simulation) Setup Used a LoRa gateway with 8 parallel demodulation paths. An equal spectrum given to LoRa and SNOW. Energy consumption and latency are quite steady in SNOW As parallelism in SNOW is much higher.

Conclusion SNOW is an LPWAN over the TV white spaces. Can be exploited by the future IoT and CPS. Can help shape and evolve IEEE 802.15.4m standard. Potential advantages over existing LPWANs. SNOW achieves scalability, robustness, energy efficiency Asynchronous, parallel Tx and parallel Rx Reliable, bidirectional Simple design using single antenna radio Future work Coexistence Mobility Security.