Pervasive Systems SD & Infrastructure.unit=3 WS2008

Transcription

1 Pervasive Systems SD & Infrastructure.unit=3 WS2008 Position Tracking Institut for Pervasive Computing Johannes Kepler University Simon Vogl

2 Infrastructure-based WLAN Tracking Real-Time Tracking of Assets / People Wi-Fi Access Points as infrastructure IEEE ; 2,4 GHz low realization effort! Tracking Methods Radio Map using Received Signal Strength Indication (RSSI) [cf. Scene Analysis] k-nearest Neighbor Pros: high accuracy, low client performance needed Cons: high calibration effort, server-side localization, server as SPOF, privacy? Signal Triangulation (Angulation, Lateration) Pros: client-based localization, no calibration needed, privacy control Cons: lower accuracy, higher network traffic, interference problem Time Difference of Arrival (TDOA) Triangulation Pros: high accuracy, less interference disrupted than RSSI techniques Cons: time measurements not supported by WLAN hardware, special infrastructure needed Hybrids cf. Pervasive Computing Infrastructure VL - Workshop Localization < 2 > Heinrich Schmitzberger

3 Available Systems Commercial Systems Ekahau Method: RSSI Radio Map Accuracy: ~1m WhereNet Method: TDOA Triangulation Accuracy: ~1m Aeroscout Method: TDOA Triangulation Accuracy: ~1m GPL Systems Place Lab Method: RSSI triangulation to access point additional usage of Bluetooth and GSM signal Accuracy: ~10m JavaWPS cf. Place Lab MagicMap Method: hybrid. RSSI multilateration, radio map with weighted graphs and statistical algorithms Accuracy: <1m (usage of localization tags) < 3 > Heinrich Schmitzberger

4 IPC Ekahau Positioning Engine: ekahau.soft.uni-linz.ac.at:8550 < 4 > Heinrich Schmitzberger

5 Ekahau Architecture Open architecture, Java based, XML API Java SDK provided Core Components Ekahau Positioning Engine (EPE) server Ekahau Location Survey (ELS) Battery powered Wi-Fi tags < 5 > Heinrich Schmitzberger

6 Ekahau Deployment Location calculations Wi-Fi tags independently listen to the network signals collect signal strength information (RSSI) relay information to Ekahau Positioning Engine data packets size: 60bytes for each location very low bandwidth usage on the network Tag features Network security features: WEP128, WPA2 and/or V-LAN, Firewall, MAC filter 2-way communication feature configure and maintain the tag over-the-air send text messages to tag s optional message display optimized to provide several years of battery life (?!) < 6 > Heinrich Schmitzberger

7 Ekahau SDK Install ekahausdk.jar (library bundle) previous to starting your sensor bundle in Equinox! mind your imports in the Manifest! String server = ekahau.soft.uni-linz.ac.at ; int port = 8550; String user = pcistudent ; String password = 3k4h4u ; public void listtags() throws EngineException { PositioningEngine engine = new PositioningEngine(server, port, user, password); List devices = engine.finddevices(); Iterator it = devices.iterator(); while (it.hasnext()) { DeviceLocation dl = (DeviceLocation) it.next(); System.out.println(dl); } } < 7 > Heinrich Schmitzberger

8 Ekahau Positioning Data Retrieval And Interpretation Tag related Data dl.getmac(); dl.getbatterylevel() dl.geteventtype() dl.gettype(); Map / Model related Data dl.getmapid(); dl.getmapname(); dl.getmodelid(); Qualitative Position related Data dl.getzoneid(); dl.getzonename(); Signal related Data dl.getquality(); dl.gettimestamp(); < 8 > Heinrich Schmitzberger

9 Determining A Near Relation Get the Coordinates of the tracked Tag dl.getx(); dl.gety(); Mind the radius 2 pixel conversion current model scale: pixel = 100 cm Calculate if tag is inside circumcircle Pythagoras!!! < 9 > Heinrich Schmitzberger

10 Useful Links Wi-Fi Location Maps / DBs Skyhook Wireless Wigle Positioning Systems Ekahau Place Lab MagicMap < 10 > Heinrich Schmitzberger

11 Outdoor positiioning MicaZ+ GPS Currently IPC: 23 sensor nodes MicaZ 9 GPS boards MTS 420 Temperature Humidity Barometric Pressure Ambient Light Sensors Dual-Axis Accelerometer GPS Module Manufacturer:

12 Quick and Dirty Computer Vision Presence detection with CMOS sensors and simple processing (Motion Detection)

13 Q: How do PIR sensors operate? PIR: Operate in the infrared spectrum Detect infrared (=heat, elmag. waves) from human bodies Detect changes from one sector to the other Works a little like an insect eye (but only 2-8 pixels) Camera: Detect visible light (elmag. waves) Compute Motion: Simple Estimator: use Pixel-per-Pixel difference to previous picture Threshold agains noise Compute optical Flow, Finding Humans???? Tricky. -> google for computer vision library Simple: Compute Regions (Morphological operators) Simpler: Detect motion in static regions Trickier: Compute Image segments,

14 Computer Vision: Code?!? Java Media Framework Access Cameras, decode and display videos Filter architecture to Example Code: X =>. Serious (research grade) code: C / C++ / DSP-Assembler code OpenCV, (google for computer vision library )

15 Processing.org Technical Sketch-pad Used to create art exhibits, simulations, visualization, Includes editor, environment Builds on Java Interfaces to HW I/O Cf. Arduino Plugin-architecture also for Video Builds on Quicktime Contributed CV libs -> p.ex. BlobDetection

16 LoD (Link of the Day): Make Magazine - Computer age do-it-yourself magazine

17

18 Next Unit Concept Presentation MAIL Your Slides (pdf) to UNTIL :00! < 18 > Heinrich Schmitzberger

19 Slide layout SLIDES MUST INCLUDE NAME; Matrikelnr in header (Your) Project title 3-Slide Version: Slide 1: Overview, approach Slide 2: Sensing Slide 3: Display 2-Slide Version Slide2: Sensing& Approach Slide 3: Display PLEASE: Export your slides to PDF Office etc: OpenOffice: Export as

20 Equipment Handout :: Web based reservation