1. A Relative Positioning System for Co-located Mobile Devices Mike Hazas, Christian Kray, Hans Gellersen, Henoc Agbota, Gerd Kortuem,Computing Department, Lancaster University, United Kingdom MobiSys’05, 2005, Lancaster University, United Kingdom

2. Outline Introduction Relate System Architecture System evaluation Conclusion

3. Introduction Relate system to enable locate mobile computing devices to establish their spatial relationship Based on wireless sensor devices dongle The Relate network is used to coordinate and collect range and angle-of-arrival measurements using ultrasonic signals between the dongles share information and uses this information to independently establish and maintain a spatial model of the network

4. Relate System Architecture

5. Relate Sensor Layer Relate Dongles - USB connector - ultrasonic transducer x 3 - PIC18F452 microcontroller - RF module

6. Relate Sensor Layer Relate Dongles - RFM operating in the 868.35MHz - RF provide communication between the dongles in the relate network - Ultrasonic 1M~25MHz low power and high frequency, can detect location of object

7. Relate Sensor Layer Relate Dongle Network - dongle devices communicate over a peer-to-peer network with random access - using TDMA technology - with 13ms packet length and 64 byte payload per packet - using seven slots for data transmission over RF and ultrasonic sensing - Relate-specific protocol function implemented on this network include Relate network discovery and management, coordination of ultrasonic sensing and collection of sensor data

8. Relate Sensor Layer Relate Dongle Network - network state as table of device IDs each with a timestamp of their last sighting - when dongle first connect to a network after interrupt protocol cycle and send request packet to sync (dongle network time)

9. Relate Sensor Layer Data collection - IDs of the transmitting dongle and receiver dongles - range (distance) and angle-of arrival estimate - receiving number of transducers - timestamp (sighting time)

10. Relate Sensor Layer Measurement data - quantitative relationships can be expressed as distance between devices - qualitative relationships describe the relative spatial arrangement of one device with respect to another (left_of,in_front_of,right_of, moving_away)

11. Relate Model Layer Relate model - The Relate model is a list of labelled graphs. - Each graph represents the spatial arrangement of devices in a Relate network in time - represent spatial relationship - share some attribute value of device

12. Relate Model Layer Data processing pipelines

13. Relate Model Layer Aggregation - the initial pipeline stage generate a new graph at the start of each processing cycle - it continuously reads sensor data packets from USB - each packet represents a measurement described as a six tuple - sensor data aggregation for the current graph until a specified condition met - condition can be time-based (every 100 data packet or every 500 ms) - when the condition is met then current graph is completed

14. Relate Model Layer Positioning - to compute quantitative attribute spatial and relation ex. position & orientation - range d ij (distance) - angle-of-arrival ψ ij cosΘ=b(xj.yi)/ dij

15. Relate Model Layer Positioning - produces new position and orientation data for all devices represented in the model

16. Relate Model Layer Abstraction - abstraction on the quantitative output of the previous stage - relative coordinates are transformed into qualitative relations that encode spatial relationships (left_of,in_front_of,right_of, moving_away) - computed from the coordinates to partition distances into categories such as nearby or far_from

17. Relate Model Layer Inference - compares the set of relations to those contained in previous graphs - generates a set of spatiotemporal relations such as approaching or moving away

18. Service Layer Event service - implements an asynchronous communication channel between the model and the application - model event when the spatial model has been change an application subscribe to events informing them to update

19. Service Layer Event service - spatial event when the spatial configuration of the relate device change - when distance to another device exceeds - network event applications can register to be notified when a device joins or leaves the network

20. Service Layer Query service - Applications can read attribute values of individual nodes or they can get a copy of the entire model - getDeviceList() returns list of all available - devices; getDeviceCoordinates() returns coordinates - of a device; and getDevicesInFront() returns devices in front of the current device

21. Application Prototyping

22. System Evaluation

23. Raw measurement error

24. System Evaluation Successful transmission rate - connectivity between the five dongles use in the experiments - a delay of about 450ms between successive transmission

25. System Evaluation Start-up delay - Twenty “start-up” experiments were performed to explore how long it takes for newly activated dongles to join the Relate network - one host device was random to shut down its dongle. The dongle was then subsequently re-started by the host - In 95% of cases, the other four dongles in the system successfully detected a radio trigger packet from the newly joined dongle within 3.7 s after it was started up

26. Conclusion sensor and model layers provide relative location and orientation estimates at an accuracy establish spatially relationship

27. Pipeline Process Job1 Job2 Job3 Job1(t)=k ． T k, Job2(t)=(n-1) ． T k Jobn(t)=(k+n-1) ． t k tktk tktk tktk

28. Service Layer Spatial communication service - can be used to disseminate data between Relate devices - getDeviceList() send(host name, msg) sends a message to a device - sendFront(msg) sends a message to all devices currently in front of the local device - sendBeyondDistance(float, msg) sends a message to all devices that are currently more than a specified distance away from the local device