Phase 1

The GreenSpace Project started as a brainchild of Dr. Thomas Furness of HITL and Dr. Masahiro Kawahata of the Fujitsu Research Institute in April 1993, and is jointly funded by the two institutions. The goals of the project are to to develop and demonstrate an immersive communications medium where distant participants feel a sense of presence in a shared virtual environment, a "virtual common". Ultimately, the project aims to promote collaboration at a distance among 100 or more participants over broadband networks such as SONET/ATM, immersed in an environment rich in visual, aural and tactile cues.

The project was divided into 3 phases. Phase 1 was the concept formation and definition, Phase 2 was a technological assessment. One of the conclusions reached in Phase 2 was that the cost of doing broadband network experiments across large distances would be prohibitively expensive at present. Additional funds and partners would be needed.

Phase 3 started in January 1994. Its purpose was to design, develop and demonstrate a preliminary testbed using conventional networking technologies. The initial testbed would be used to explore the feasibility of the GreenSpace concept, to provide visibility to the project to aid in raising additional funds and to find additional partners.

Phase 3 culminated in a public demonstration linking Tokyo and Seattle in a realtime VR teleconference during November 14-17, 1994. The demo site in Tokyo was the NICOGRAPH Convention (the Japanese equivalent of SIGGRAPH), and the Seattle site was the HITLab.

The Demo

Key features of the demo at NICOGRAPH Convention were:

  • Low bandwidth networked VR using multiple ISDN BRI lines.
  • 2 participants on each side of the Pacific, local participants faced each other at a table. Remote participants appeared to the side.
  • Participants' faces were scanned on the spot and sent to the other side, so they could see each other. Initially it was planned to select different facial expressions based on the speech data.
  • The US participants saw a Japanese meeting room, with Mt. Fuji in the background.
  • The Japanese participants saw a western log cabin with a view of Mt. Ranier.
  • At each corner of the table was a goal post; the 4 participants engaged in a cooperative game of herding creatures into these goals, complete with sound effects.
  • Participants can hear each other, albeit poorly.

    The hardware configuration used for the demo is shown below:

    Click for large version

    Network Description and Bandwidth Requirements

    The trans-Pacific network for this demonstration is illustrated in subsequent diagrams.

    Click for large version

    Since we had two participants on each side of the Pacific, we needed to track the motion of each participant's head (for viewing) and hand (for interacting with the world). This meant we sent information on 8 tracked objects across the network. The payload for doing this is:

    (8 objects) x (6 DOF) x (32-bit floats) x (30Hz) = 46Kbps
    Typically, the loss of bandwidth due to IP encapsulation overhead and packet collision reduces the payload to about 70% of the line bandwidth. This means that 8 tracked objects required approximately 46Kbps/0.7 = 64Kbps.

    Since we also wanted one mono audio channel to be transmitted bidirectionally over the net, the payload would be and additional 2 x 8KHz x 8-bit ulaw samples = 128Kbps. This is telephone quality audio. Again, taking into account IP overhead, the required bandwidth becomes: 128Kbps/.7 = 183Kbps.

    Thus, for a simulation with 2 participants + 1 mono audio channel on each side, our expected bandwidth requirements were approximately:

    183Kbps + 64Kbps = 247Kbps

    Click for large version

    The ISDN BRI (Basic Rate Interface) has 2 B channels for data and 1 D channel for signalling and control. Each B channel can carry 56Kbps or 64Kbps, depending on the carrier. Assuming we can get 64Kbps per B-channel, we need 2 BRI lines to support the 247Kbps requirement above.

    Using a bandwidth-on-demand ISDN/Ethernet bridge or router (such as the Ascend Pipeline 400B), B channels can be combined to form one logical channel with higher bandwidth. According to the rough calculations above, we required 4 or more B channels for the demo. To be safe, we used 6 B channels (3 BRI lines) for the IP traffic.

    We also used a conventional video teleconference system so that the crowds can see each other. The teleconference system required one additional BRI line.

    Phase 2

    The objective of GreenSpace II is to apply the concept of distributed VR to a particular application domain. The chosen application is architectural design. Architectural design requires a great deal of coordination and communication between different individuals. Architects need to communicate design ideas between themselves and between others. They need to communicate to their clients, engineers, building officials and many other types of individuals. To communicate architects traditionally use drawings and models, as well as normal business communication devices such as the fax machine and the telephone. A common situation, where architects present their work, is called the design review or design presentation. In academic settings this occurs when a student presents his design work to group of critics, and in a professional setting it occurs when an architect presents his design ideas to a client. The GreenSpace II efforts have focused on this common situation of a designer presenting his work to a particular audience for analysis and commentary .

    GreenSpace Environment

    In March of this year the GreenSpace II technical team working with the CEDeS lab produced a working prototype for distributed design reviews in virtual space. To demonstrate this system the CEDeS lab selected a simple architectural problem the design of a hotel Guestroom. This program was selected because it was small enough so as not to require too complex a model for each design proposal, but yet had enough potential to provide dramatic differences in spatial quality. Three different guest rooms were designed. For each room different versions were generated to show the room as an empty space, and as a furnished room.

    In addition a number of tools were implemented which would facilitate a remote discussion of the architectural space. Each participant is represented as an avatar with a stylized hand/pointer. In each alternative guest room there was a work table which had a scale model of each of the design proposals. This allowed the participants to view each of the three alternatives while there were immersed within one of the rooms. To support the architectural discussions of the participants each of these scale models could be "clipped" with a clipping tool. This provided a way to "clip" the models in order to see a plan or section of any one of the designs.

    On the table were place additional icons which represented different versions of each of the three guest rooms. In the first demonstration of the system each of the three rooms had three associated icons. One of the icons represented that room rendered in a neutral color and flat shaded. The second icon represented the room rendered with textures. These textures were generated by the radiosity application Lightscape, and represented a specific lighting condition. The third icon represented the room with a furnished and illuminated condition. The current system allows for two additional room alternatives which will be variations in furnishings, color and/or lighting schemes. By selecting any one of these icons a participant could move through a jump-portal to that version of a particular room.

    The CEDeS lab and the GreenSpace II team will be carrying out a number of experimental design reviews with invited architects from the Seattle professional community and the Department of Architecture. Currently the system is distributed within the HIT Lab itself. It is planned that future work will include a geographically remote distributed implementation of the system.


    See Also:

     Greenspace VRML Archive