"Thoreau: An Experimental, Low-Power Wireless Underground Sensor Network For Soil Sensing"

Transcription

1 "Thoreau: An Experimental, Low-Power Wireless Underground Sensor Network For Soil Sensing" Xufeng Zhang, Argonne National Lab Arseniy Andreyev, U Chicago Monisha Ghosh, U Chicago (monisha@uchicago.edu) Supratik Guha, U Chicago & Argonne National Lab Acknowledgements:

2 Topics The agriculture IoT use case. Wireless needs of agriculture. Thoreau: the man, the experiment. Challenges, solutions, next steps.

3 Agriculture: unique challenges for IoT Enormous potential to benefit from the huge advances in computing, communication and sensing. Requirements are quite different from other IoT environments. Zone of interest: root zone, 6 12 underground. Fields are heterogeneous: sensing needs to be spatially dense. Installations cannot interfere with normal field management practices: no antennae sticking out of the ground, low-power. Tremendous savings (2x 5X) possible in water, pesticide and fertilizer use, with important consequences for the environment.

4 Agriculture Vision: Cyber Physical System Continuous soil monitoring using dense sensors, combine with weather and crop data in the cloud and close the loop with recommendations.

5 Thoreau: the man, the experiment Henry David Thoreau Essayist, naturalist, environmentalist. Believed in recording the environment to better understand it. The experiment: First functioning buried wireless sensor network on an university campus that collects and curates time and geotagged data on an open platform on the cloud. Completely buried sensors, radio and antenna. Wireless channel Sensor hardware

6 Wireless challenges Transmission through soil depends on: Frequency (lower is better, but antenna size increases). Soil composition and bulk density. Volumetric water content: wetter is worse. Sigfox Approach In the 900 MHz unlicensed bands. Frequency hopped, narrow-band, with combining in the back end Low data rate: 12 Byte packets Simple protocols: mostly uplink. Designed for long range

7 Wireless Network Architecture Antenna Sensor Sensor Sensor Sensor I2C UART SDI Analog GPIO MCU Radio Power Management Battery Filter Amp Base station Backend Internet Callback API User Interface Sensors Nodes Base station Backend Users

8 Sensor Node Various sensor types and data protocols Sensor Sensor Sensor Sensor I2C UART SDI Analog GPIO MCU Antenna Radio Power Management Battery Filter Sigfox backend Amp. Internet Base station User interface Sensors: Temperature Soil moisture Soil conductivity Water potential 5-year operation

9 Sensor Types Decagon GS3: Temperature: -40 to 60 o C (res. 0.1 o C) Electric conductivity: 0 to 23 ds/m (res ds/m) Soil moisture: 0 to 100% (res. 0.2%) Decagon MPS-6: Temperature: -40 to 60 o C (res. 0.1 o C) Water potential: -9 to -100,000 kpa (res. 0.1 kpa) DS18B20: Temperature: -55 to 125 o C (res. 0.5 o C)

10 Sensor Power Management Power consumption: MCU: active 0.29 ma, sleep 2 ua. Negligible. Sensors: 35 ma active (0.15 sec), 0.33 ma standby. Relatively small. Radio: TX 110 ma (1 sec). With our hardware power management solution standby current~ 60 ua. Small 6 x 3 AA batteries (@4.5 V; ~6000 mah considering discharge effect 5-year operation

11 Sensor Installation

12 6 underground 1 antenna on top of ERC building 27 sensor units in service all over the campus. Locations: Logan center (Clay. 2 units) Mansueto library (sand) Alumni house (mixed) Haper quad (mixed) Ida Noyes court yard (mixed) Graham School (mixed) Kenwood (mixed) Two sensors on each unit, measuring 4 soil properties: soil temperature soil moisture electric conductivity water potential

13 Wireless Transmission Through Soil d 2 θθ 2 h 2 d 1 θθ1 h 1 L 1 L 2 Total channel loss: * ρ bk : bulk density ** soil composition in wt% Soil samples Interface: Aboveground: Underground: Determined by soil properties

14 Results ID Location Soil type Underground Depth RSSI (dbm) SNR (db) Delay (s) 5784 Logan center 1 Clay Logan center 2 Clay Alumni house Mixed Mansueto library Sand Haper quad Mixed Ida Noyes Mixed Graham School Mixed Kenwood Mixed Minimum measured signal: SNR 2.46 db, RSSI -139 dbm (0.6 mile, 10 underneath dense clay) Farthest measure result: 1.4 miles away, antenna on the ground, RSSI -117 dbm Sigfox: Uplink TX power 26 dbm, typical link budget around 160 db Rep. 160 Discrepency Calc. path loss Path loss (db) Logan W Logan E Kenwood Graham Ida Noyes Alumni Harper Mansueto

15 Preliminary Sensor Data Temperature (degree C) period: 1 day Signal strength (dbm) Raining 12 10/4/ /6/ /8/2016 Time 10/10/ /12/ /14/ /4/ /6/ /8/2016 Time 10/10/ /12/ /14/2016 Volumetric water content (%) /14/ /16/ /18/2016 Time Sprinkler on 10/20/ /22/2016 Temperature (degree C) /14/ /16/ /18/2016 Time 10/20/ /22/2016

16 Heat Maps Signal strength map Signal-noise ratio map Temperature map Electric conductivity map Soil moisture map Water potential map

17 Packet Error Rate Performance Ricean fading is the best fit for the composite channel: underground and over the air

18 Challenges and solutions Challenge: Packet loss rate rises with increasing depth Potential Solutions: Increase transmit power Diversity combine at receiver Use lower frequency: TVWS a possibility Improve antenna type and orientation Future Work Add more soil sensors: nitrates, biological, ph etc. Deploy LoRa, for comparison. IEEE ah is also a possibility Large scale experimental deployment under discussion with the Morton Arboretum and Chicago Tollway.

19 Conclusions Proof of concept demonstration of fully buried, dense sensor network at 900 MHz, from soil to cloud. Low- Pow er Cloud Curation+analytics+physics Cn Soil Sensing PM 2.5 Sen sor s sensors Power wireless Data rate Range