Engineering and Augmented Reality Tour Guide

We describe a mobile augmented reality system intended for in situ reconstructions of archaeological sites. The evolution of the system from proof of concept to something approaching a satisfactory ergonomic design is described, as are the various approaches to achieving real-time rendering performance from the accompanying software. Finally, some comments are made concerning the accuracy of such systems.

Transcript

1. Engineering an Augmented Reality
   Tour Guide
   Panagiotis D. Ritsos, David J. Johnston, Christine Clark
   and Adrian F. Clark
   VASE Lab, Department of Electronic Systems Engineering
   University of Essex
   IEE Eurowearable, Birmingham, 2003
   

   View Slide

2. Talk Outline
   Augmented Reality in General
   Project Definition
   1st Generation system
   Limitations and Solutions
   2nd Generation system
   Conclusions
   

   View Slide

3. The Concepts
   Augmented Reality
   An augmented reality system superimposes virtual information,
   usually computer generated models, on top of the real
   environment in order to enhance the latter. The user sees a
   composite image of the real and the virtual world.
   Context Awareness
   “any information that can be used to characterise the situation
   of an entity, where an entity can be a person, place, physical
   or computational object” - Dey & Abowd (1999)
   

   View Slide

4. The Application
   Augment the Gosbecks Temple en situ using a 3D
   reconstruction of the temple
   Use a Wearable Computer to
   Interface with a GPS and an HMD tracker to extract
   position and orientation information
   Render the model according to this information
   Use a Distributed instead of a Centralized system
   No centralized rendering – each wearable renders its own
   view
   This enables more wearables to roam in our virtual space
   Eliminates the overhead of transmitting the 3D information
   to each wearable
   

   View Slide

5. The Virtual Temple
   The Gosbecks Park and the Temple to Claudius
   We created a 3D Model of the temple using OpenGL
   

   View Slide

6. The Tour Guide in a few words
   The user roams with the
   wearable on the field
   Position extracted from the
   GPS unit
   Orientation extracted from the
   HMD tracker – Context
   Awareness
   A virtual model of the temple
   is projected through the HMD -
   Augmented Reality
   Gosbecks is ideal for such an
   application because it is
   relatively flat and with no
   surrounding buildings – the
   GPS unit can ‘see’ a large
   number of satellites
   

   View Slide

7. Wearable Computers
   A Wearable computer is
   usually belt-worn or carried
   in some form of a jacket
   vest or backpack
   It has the processing power
   of a modern laptop
   Components usually include
   a Head Mounted Display,
   and Input device such as a
   ‘chord keyboard’
   Extra in most Mobile AR
   applications is a Global
   Positioning (GPS) Unit
   Camcorder batteries used
   for power
   

   View Slide

8. Remus Wearable – 1st Generation system
   Pentium 266, 64 MB of
   RAM, generic VGA,
   4 serial ports, USB,
   Audio and PC Card adapter
   to be used with WLAN cards
   Constructed from PC/104
   cards
   All encompassed in an
   aluminium box
   Garmin GPS unit,
   Orientation Tracker on the
   HMD
   4.9 Kg and operational for
   45 min
   

   View Slide

9. Limitations of initial configuration
   Very Slow refresh rate – No 3D acceleration
   Accuracy of GPS
   Bulky system – can be tiring
   Power supply is not adequate –2 hours of operation are
   required
   Requires a simpler interaction mechanism than a chord
   keyboard
   

   View Slide

10. GPS – Limitations
    It is not accurate enough –
    Drift with time
    Differential GPS could
    improve performance but
    not to the desired level
    Ideally an accuracy of 20
    cm (position) is required for
    ‘True’ AR
    Orientation requirements
    not so stringent – about 5o
    error is acceptable
    

    View Slide

11. Proposed solutions
    Computer is integrated to a
    vest
    Optimisation of the model
    Use View Frustum
    Culling to render only
    visible objects
    Employ levels of detail
    processing – replace
    distant objects with
    images
    Localisation? GPS is
    adequate for a
    demonstration but not for a
    commercially viable system
    

    View Slide

12. View Frustum Culling
    A technique used in 3D games like First Person Shooters
    (Quake, Unreal etc.)
    OpenGL renders all objects requested – even those that are not
    always visible!
    View Frustum Culling enables us to ‘cull’ objects that are
    not visible
    

    View Slide

13. Levels Of Detail
    Simple mechanisms for further
    optimization
    Distance Test – If something is
    very distant, do not draw it! –
    In the colonnades we draw one
    row (front) instead of two
    If Something is moderately
    away, draw it without detail –
    This reduces polygon count
    significantly since each column
    has at least 4 extra cylinders
    

    View Slide

14. Romulus Wearable – 2nd Generation
    System
    Mini – ITX based system
    Faster, simpler, extra
    multimedia features,
    comparable power
    requirements –
    increased fun factor!
    Large upgrade potential
    Large support from online
    community due to platforms
    popularity
    And recently … 3D
    hardware support!
    Integrated into a
    Photographers Vest – easier
    to wear and carry
    

    View Slide

15. Romulus Wearable – Part 2
    Via C3 1Ghz Nehemiah
    VIA ProSavage CLE266 chipset
    512 MB DDR memory
    20 GB Hard Disk
    Soundcard, Ethernet 802.3, 2 USB, 2 Serial, 1 I2C
    Aluminium casing – compatible with all mini-ITX modules
    Garmin GPS Unit, Virtual I/O HMD with orientation tracker
    Netgear 802.11b wireless LAN interface
    Comparable Power requirements to Remus, with 2A max drain,
    1.3 average
    System powered from a custom battery pack of 10 1.2 Volt
    NiMH cells – two pack to be used in finalised system.
    

    View Slide

16. System Performance Evaluation
    System Testing is currently performed
    Graphics speed significantly better – 7 frames per second
    average, 15 fps maximum.
    System is much lighter than alternatives, yet its performance
    as far as graphics are concerned is more than adequate for
    demonstration purposes
    Next step is to improve the interaction mechanisms – use a
    single button interface to restart the application
    The system has increased upgradeability potential than other
    embedded boards alternatives – online community provides
    significant feedback – drivers constantly updated.
    Yet…ideally it should be smaller!
    

    View Slide

17. Conclusions
    Augmented Reality prototypes provide adequate
    demonstration systems – 3D rendering capability and
    GPS accuracy the major drawback
    Commercially viable systems require more accurate,
    smaller, faster and lighter wearable computers
    Hardware implementation of such a system is feasible –
    similar to a laptop with good multimedia features
    Localisation techniques require further investigation
    Power consumption problems similar to modern laptops
    

    View Slide