1. airMessages interactive density exploration Steven Strachan Hamilton Institute Roderick Murray-Smith University of Glasgow and Hamilton Institute http://www.dcs.gla.ac.uk/~rod

2. We introduce here a system that offers a general mechanism for providing highly interactive context- aware applications. In our example application we use the system to guide users to messages left in the local augmented virtual/physical environment. This system may be used in a more general way to allow users to probe and explore local contextual areas of interest.

3. Mobile Sensing: MESH Modality Enhancing Sensor-pack for Handhelds Designed for the IPAQ range of pocket PCs Physical design same as the PCMCIA expansion jacket Triple-Axis acceleration sensing –MEMS Accelerometers –Orientation sense and gesture capture High Fidelity Vibrotactile Display –Sample based, Non-Volatile sample storage,Audio bus-driven option –Actuator – VBW32 rewound Triple-axis magnetometer –Orthogonal Magneto-Resistive elements Capacitive sensing GPS

4. Next sensor pack.. SHAKE Our next generation pack… Bluetooth, wireless and compact Accelerometers, gyros, magnetometers and haptic feedback. Use for head, device or bimanual gestures.

5. GPS Navigation Problem GPS is useful but inaccurate Inaccuracy varies in a complex way –Reflections, shadowing, poor coverage –Could use hybrid positioning General problem – accurate representation of belief and trustworthiness

6. Uncertain Display Poor displays lead to poor control Norman's example of The Royal Majesty “precise” position

7. Uncertainty in GPS Navigation Represent and display the true uncertainty of the navigation system – make it “honest” realistic display should regularise control behaviour Incorporate models environment models user models Monte Carlo sampling is a convenient statistical technique for dealing with uncertainty

8. GPS Model Uncertainty in location –Affected by nearby occlusions –Also reported from GPS device Can model areas of coverage with shadow maps [Steed 2004] –Trace visibility of satellites given a known map of buildings

9. Shadow Map Example –Raytraced from satellite positions –Darker regions indicate areas of lower coverage

10. User Behaviour Model user as dynamic system –Heads in current direction (e.g. from magnetometer heading) –Flows around obstacles Gradient following model is simplistic but sufficient.

11. Complete Model Represent current user position as samples from a Gaussian distribution –Mean at current GPS location –Variance determined by shadow mapping Propagate those samples through a simple gradient following model –Display resulting particle distribution at some time horizon

12. the major benefit of this work is the introduction of a new kind of rich, embodied and location-aware spatial interaction with the environment. A rich and embodied interaction enables a user to interact and traverse densities which they place over the real world.

13. This system utilises both audio and vibrotactile feedback. The audio feedback consists impact sounds generated using granular synthesis –…these impacts are also mapped into vibrotactile feedback. Feedback…

14. System Testing Initial testing was conducted to demonstrate that users could find and leave messages in their virtual environment. –6 participants Participants followed a set scenario around an area of the campus

15. All participants successfully completed the task required of them.

16. Not all participants really used the interface to its full potential.

17. Conclusion Experimental participants performed well and had a natural intuition for the task, which meant that learning was quick. It is hoped that this system will aid the creation of a new kind of location-aware computing, one which allows a rich and embodied spatial interaction with the local environment. We have demonstrated that the probabilistic, negotiated interaction techniques can be applied effectively to the mobile GPS navigation problem.