IOT GEOLOCATION NEW TECHNICAL AND ECONOMICAL OPPORTUNITIES

Transcription

1 IOT GEOLOCATION NEW TECHNICAL AND ECONOMICAL OPPORTUNITIES Florian LECLERE EOT Conference Herning 2017 November 1st, 2017

2 AGENDA 1 NEW IOT PLATFORM LoRa LPWAN Platform Geolocation service BUSINESS OPPORTUNITY 3 2 FIELD TESTS EXPERIMENTATION Algorythms Static / Walking scenario Results Improvements Q&A 4 2

3 1 NEW IOT PLATFORM : LPWAN / LORA

4 WHICH TECHNOLOGY FOR WHICH PURPOSE? Local networks o Short range o Low autonomy LPWAN networks o Long range o High autonomy o Low data volume o Indoor & outdoor o Free and open-access frequencies o Low installation costs GSM LPE cellular networks o For large data volumes o Low autonomy o High operating costs

5 TYPICAL LORA NETWORK ARCHITECTURE Applications Smart cities Smart agriculture Santé Smart Grid Smart metering Tracking IoT Operators Objects connected via sensors Stations Core network 5

6 AT THE CENTRE OF THE LORA ECOSYSTEM LORA ALLIANCE In members In 2017 Already over 500 members 100 Roll out 6

7 FOCUS ON THE INFRASTRUCTURE ECOSYSTEM A WIDE RANGE OF SERVICES MANAGEMENT OF ROLL-OUT MANAGEMENT OF THE NETWORK MANAGEMENT OF CONNECTED OBJECTS SUPERVISION Management of planning process Radio Planning Management of stations Network management Data routing Management of connected objects Advanced geolocation services

8 GEOLOCATION OVER LORAWAN NETWORK 8

9 LPWA NETWORK LOCATION BASED SERVICES (LBS) OMC Operation and Maintenance Center IoT Location Base Service Located by the infrastructure No dedicated devices/features, no aiding Low Power/Cost No extra cost for location feature High sense communication means long range => low cost infrastructure Technology (LoRa TM ) makes it possible Wideband modulation, geo-redundancy, network topology, unlicensed band Low cost communication technology chipset technology, volume and Radio foot print

10 SOURCE LOCALIZATION Methods: Range-based uses the absolute distance estimates or angle estimates in order to calculate the location (ex. RSS, AOA, TOA, TDOA) Range-free requires neither distance nor direction measurements (ex. Nearest neighbor, fingerprinting & machine learning, )

11 RANGED-BASED POSITION ESTIMATION ALGORITHMS Algorithm Type Advantages Disadvantages Linear Least Linear Global solution guaranteed Low accuracy Square (LLS) Computationally efficient Noise statistic not needed Linear Weighted Linear Square (WLS) Linear Global solution guaranteed Higher accuracy than LLS Noise statistic needed Nonlinear Nonlinear Least Square (NLS) Nonlinear Higher accuracy than linear approaches Noise statistic not needed Global solution not guaranteed High complexity Maximum Likelihood (ML) Nonlinear Highest accuracy High complexity Global solution not guaranteed Noise statistic needed

12 KERLINK S SOLVER Range-based Method: TDOA (fine stamping) Multi-algorithm: NLS & ML Timestamp noise estimation (LOS case) Pre- and Post-processing

13 SOLVERS AND PROCESSING FILTERS Kerlink solution integrates different algorithms to fit any use cases : - Static End-point - Moving End-point Post-processing filters are also provided to avoid aberration point and improve accuracy : - Kalman Filter - Correlation with previous estimation - Information of number of GW receiving the messages

14 KERLINK S SOLVER All base stations share a common time base (e.g. provided by GPS or similar) and requires a new High resolution time-stamping in Hardware Packet received by multiple base stations Each base station reports the time of arrival & other data such as signal strength, signal to noise ratio etc. At least 3 gateways are required 4 gateways recommended 14

15 LOCATION BASED SERVICES: 1- Applicative Radio datagram Demodulation accuracy multipath Clock Synchronisation LBS 2- RX parameters RX signal strength Noise level Time of Arrival Estimated system noise 3 - Range based solver Correlation of Time Differential of Arrival Environment modeling Propagation behaviors

16 Cost Adding GPS Module, Antenna, battery => about at least >$5 LPWAN GPS VS LBS GPS CONSTRAINTS Power consumption From off state, getting a position costs: Full GPS x300, offboard GPS processing x30 Sky view GPS requires a stable view of satellite constellation => installation constraints Applicative payload Means on applicative security context 16

17 2 FIELD TESTS EXPERIMENTATION

18 LEGEND STATIC POINT Results from 22th September suburban area Gateway location Real Device Location Average location calculated over 20 min Result displayed with CDF accuracy for 30% - 50% - 67% - 90% Average error rate WALK TEST Results Points location computed by Algorithm Walk Test points Average Error Rate

19 SCENARIO STATIC POINT Below you will find the results display on Google Earth : Results Points location calcultated by Algorithm Gateway location GPS Points Real Journey Coverage & Interpretation Distance between GW is 1,5km Suburban radio condition Average error 60m Precision (m) CDF (%) 30% 50% 70% 90% 95%

20 SCENARIO WALK TEST Below you will find the results display on Google Earth : Results Points location calcultated by Algorithm Walk Test points Coverage & Interpretation Distance between GW is 1,5km Suburban radio condition Average error 100m One message sent every 30s

21 COVERAGE IMPACT We tried to investigate on the impact of number of message taken into account for a position calculation. We took static analysis for Device 1. Depending of message, they could be received properly by 3 to 8 stations Average error (m) Number of station coverage 82,1 3 79,1 4 90,5 5 88,4 6 67,6 7 95, ,0 100,0 80,0 60,0 40,0 20,0 0,0 Average error on various coverage Average error (m) ANALYSIS : Having an endpoint cover by several stations allow to give more weight to the geolocation result but it does not seems interesting to eliminate points when received by a limited number of stations 21

22 CDF RSSI IMPACT We decided to enhance the level of the RSSI filtering threshold on the solver for the Device 1 Static analysis The main objective was to see if the reduction of points computed by the solver giving them strengthen signal would give a better accuracy. Analysis: First approach shows that this assumption was not that relevant on the accuracy. RSSI/SNR is linked to multi path Better accuracy in sub-urban area 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% Dev1 - Static Pre-filter SNR<-25dB Pre-filter SNR<-19.5dB Error (m)

23 DOP (DILUTION OF PRECISION) IMPACT The Dilution of Precision (DOP) is an indicator to evaluate the positioning accuracy related to the geometrical location of gateways and endpoints. We investigate DOP Impact on the average error We took the Device 1 Static analysis Average error (m) DOP 83,1 < 0,8 83,5 < 1 82,8 < 3 83,2 No limit ANALYSIS : At this stage, there is no obvious correlation between the DOP and geometrical position of stations on the accuracy results. Nevertheless, the DOP analysis and calculation is one of the main axis for our accuracy improvement 84,0 83,5 83,0 82,5 82,0 Average error on various DOP < 0,8 < 1 < 3 No limit Average error (m) 23

24 SOLVER IMPROVEMENT INVESTIGATION To guarantee an accuracy improvement, Kerlink teams are working on : Work and calibrate on the spatial diversity space (antennas) and the time diversity (packets) Detect and eliminate false and aberrant (eg NLOS and multi-path cases) Estimation and modeling of channel and noise Improvement of Kalman filter DOP weight and analysis Objective is to reach 30m precision in the next months

25 3 BUSINESS OPPORTUNITIES

26 Birth of the internet 25 years ago OVER 15 BILLION OBJECTS CONNECTED IN 4 YEARS 1 billion websites (1) 3.2 billion web surfers (2) Smartphones 9.2 billion (3) From 15 (4) to 200 (5) billion connected objects (6) by 2020! (1) Internet Live Stats 2014 (2) International Telecommunications Union (3) Ericsson Mobility Report, November 2015 (4) Machina Research 2015 (5) IDC: 212bn objects connected by 2020 (6) Excluding tablets and smartphones

27 LOCATION BASED SERVICES: WHICH TYPICAL USES CASES? Object tracking (LBS): Focused on status (IN/OUT/MOVING) Long lifetime expectation (>5Y) Less constraints for accuracy Strong constraints on device cost ASSET AND INVENTORY MANAGEMENT Mobile tracking (GPS) Focused on route estimation External power supply (or release lifetime constraints) allows accuracy Frequent position (short tracking) AUTOMOTIVE LPWAN network management Performance management (radio monitoring) locate once

28 LOCATION BASED SERVICES: LOCATION CENTRIC MARKET ANALYSIS GPS LBS In 2020, global revenues in B (*) Emergency Defense (40) Automotive (600) Smart city Transportation (450) Connected health (1000) Supply chain Tracking (350) Retail goods (100) Smart Building (110) Application oriented Agriculture (160) Smart Metering (1300) (*) Source: Machina Research

29 LOCATION BASED SERVICES: LOCATION CENTRIC MARKET ANALYSIS (2) GPS LBS In 2020, global revenues in B (*) Smart City Public transportation (450) Electrical Vehicle (20) Public Transportation (20) Environment (40) Road traffic mngt Smart Lighting Waste mngt (370) Stolen Vehicles (60) Asset Tracking (150) (RFID, Inventory) Supply chain (40) Warehousing storage (20) Supply chain Tracking (350) + Manufacturing/vending machine (*) Source: Machina Research

30 Easy to deploy Any endpoint can be located once being deployed, small location beacons can be designed Energy saving no extra power consumption - no GPS -, end device is located when it communicates Innovative LoRaWan feature - part of the network Customers benefits Cost efficient No additional hardware or software cost, no impact on device BOM - no GPS -, no Capex

31 1 Available ASSET Tracking applications On IoT platforms 2 Infrastructures and services ready LBS Sensors ready Validated Business models MARKET IS READY SEVERAL INDUSTRIAL PRIVATE NETWORKS IN DEPLOYMENT 31

32 THANKS TO 32

33 QUESTIONS? Florian LECLERE