laboratory for communications engineering engineering department, university of cambridge location...

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Location of Mobile Devices Using

Networked Surfaces

James Scott Frank Hoffmann jamesscott@acm.org fh215@cam.ac.uk

http://www-lce.eng.cam.ac.uk/

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Overview

• Quick intro to Networked Surfaces

• Location process

• Simulations, measurements and visualisations

• Improving accuracy

• Applications

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Networked Surfaces Concept• Provide network connectivity using physical surfaces

• Such as desks, floors, etc.• Make use of gravity

• No “plug”; no special position/alignment required• Provides mobility for devices• Offers transparency of connection for users

• Support a range of services• Ethernet-style inter-computer networks• Slower serial busses for peripherals• Power

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Networked Surfaces Implementation• Augment surface and objects with conductive pads

•Different object “footprints” guarantee different numbers of channels

• When connecting, “pad mappings” are discovered

• Prototype characteristics:– PCMCIA interface to notebooks – Connection in ~0.2s– Disconnection in ~0.1s– 5Mbit/s networking

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Object Pad Configurations

Links Required

Object Pads

Footprint Diameter

(mm)

2 5 26

3 9 46

4 12 68

5 16 88

6 19 110

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Location Process

Object Placement

Surface Strip Object Pad

3

16

5

10

7

4

Pad Mapping

LocationAlgorithm

Object Location

y

x

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Location Algorithm

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Location Characteristics

• Location available for 100% of connected objects

• Expect guarantee of bounded maximum error

• Algorithm is fast: ~1ms on modest hardware

• Tested using simulations, measurements and visualisation…

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Simulations• Simulation process:

– Simulate random placement

– Calculate pad mappings

– Execute location algorithm

– Compare result with original placement

• Allows fast testing of many placements– 1,000,000 locations tested for each data point

• Other advantages– Testing of various footprints

– Evaluation of possible improvements

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Simulation Results

17mm

61mm

10°

41°

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Comparison with Measurements

• 50 manual measurements

• 4 link object

• Est. 5mm accuracy

• Results very close to simulation

Variable Mean Simulated Error

Mean Measuremen

t Error

Difference

X 15mm 13mm 2mm

Y 3.6mm 3.0mm 0.6mm

(X,Y) vector

16mm 14mm 2mm

7.8° 6.3° 1.5°

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Visualisation Tool• Circle shows est. position,

rectangle shows bounds

• Lines show est. orientation and max orientation range

• Y accuracy >> X accuracy

• 2 column accuracy >> 1 column accuracy

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Improving Location Accuracy

• Current prototype does not provide full pad mapping info• Only as many links as necessary, and only one object pad per link

• Can augment with information on “Duplicate Pads”• For each surface pad used, list all object pads touching it (instead of just

one)

• Can also augment with information on “All Links”• Provide mappings for all surface pads sensed, not just those used for

connection

• Possible to implement in current prototype• Changes only required in FPGA programs, not in hardware• Use simulation to test improved performance

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Improved Simulations — (X,Y) vector

17mm

61mm

8mm

32mm

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Improved Simulations — Orientation

9°2°

10°

41°

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Integration and Applications

• Integration with context-aware middlewares– E.g. QoS DREAM Flame, SPIRIT (both at LCE)

• APP: Auto-configuration of devices– Automatically connect devices appropriately– e.g. keyboard connects to closest monitor

• APP: Interface mobility– Remote interfaces using devices with better I/O hardware– e.g. ad-hoc docking station for a notebook computer

Laboratory for Communications EngineeringEngineering Department, University of Cambridge

Conclusions

• Networked Surface prototype is capable of locating devices with a mean error of 8mm and 2º

• Also guarantees maximum errors of 32mm and 9º

• Beats most “dedicated” location systems!

• Many useful applications, including surface-centric ones