AVID -- Q/A

in response to reviewers' comments

dr. A. Eliens 
18/04/02


Q1. Is AVID only an engineering project? (#2)

A1. No, of course not! The research theme falls within the
    more generic theme of 'intelligent multimedia' and the project
    attacks both conceptual and technical issues in the more
    delimited area of agent-based virtual reality.
    As argumented in [3], we regard desktop virtual reality
    as a potentially powerful interface for multimedia information
    systems.
    The WHY of our research may be indicated by words such as
    'adaptive' and 'flexible'. The HOW by pointing to conceptual
    issues, as for example the scripting language based on
    dynamic logic described in [2]. See the additional references
    below.

Q2. What ACL are you planning to extend? (#2)

A2. This is part of the research. We need to develop a suitable
    syntax and semantics that fits within our framework. A comparative
    evaluation of existing ACLs is planned for in the project.

Q3. DLP is a first-order LP language.
    How do you plan to support BDI models? (#3)
   
A3. DLP is an object-oriented first-order LP language. It has more
    powerful expression capability than ordinary LP languages. We will
    develop a general framework to simulate the modal operators
    as object components in DLP.
    As we have indicated in the proposal, we already extended
    the BDI model with so-called sensors and effectors, to enable
    agents to interact with the world. As mentioned in A1, the
    DLP-based platform allows for extensions including dynamic
    logic and also more 'exotic' logics as well as 'reflection'.
    The M.Sc. work done by our candidate researcher gives
    an indication of how we will approach the 'mental' capabalities
    of our agents.
    In the existing work of WASP, like WASP soccer games, the
    mental attitude components are implicitly constructed in the
    agent architecture. In AVID project, we plan to develop a general
    kernel of BDI agents in which the mental attitude components are
    explicitly constructed.

Q4. Why do you focus on VRML, instead of X3D and Java3D? (#3/4)

A4. Our focus on VRML is motivated by pragmatic reasons. However,
    as indicated in [2] (see references below), we are in the
    process of migrating to X3D. Java3D is not a viable option,
    since we aim for web-based virtual environments. Besides,
    our contacts within the Web3D organisation indicate that
    X3D will subsume Java3D.

Q5. Can a Ph.D. student do all the work proposed? (#3)

A5. Yes, that is to say that both Eliens and Huang will actively
    participate in the research. 

Q6. Does AVID improve the state of the art in 3D agent-based virtual worlds? (#4)

A6. Certainly. By providing a unified framework we may improve
    on existing systems. Of course, we need to learn from
    the achievements of other groups, in particular with respect
    to (behavioral and emotional) models. But then again our
    platform allows for a novel approach in these areas, as
    for example the use of dynamic logic to realize the
    STEP scripting language described in [2].

Q7. Why do you use DLP? (#4)

A7. Despite the fact that DLP exists for over a decade, its
    implementation is relative free of legacy solutions.
    It is a powerful language with a strong semantics, and
    it is based on the Java platform, which allows for easily
    incorporating a variety of functionality,  shielded
    by DLP object interfaces.
    As mentioned in A4, we are in the process of migrating to
    X3D. More in general we are working on XML processing
    capabilities in DLP. A reference to 
    encoding behavioral scripts in XML is given in the section
    on additional information, below.


Q8. What happens when the blaxxun browser becomes obsolete? (#4)

A8. Indeed, blaxxun is in problems. However, the blaxxun browser
    source code has been donated to the Web3D consortium, of which
    the PI is a professional member.
    When the browser actually becomes obsolete, we simply migrate
    to another browser, which may be either Cortona,
    the Open Worlds browser, or even the Java3D-based open source
    Web3D browser.
    Since we use the standard Java EAI this migration should be
    relatively painless.  
    Even the new X3D Scene Authoring Interface can be easily
    accomodated.
    We will however only undertake the actual migration
    when it becomes necessary.

Q9. Will the platform be built from scratch? (#4)

A9. No of course not. Our work has aimed at building a platform
    which allows for reusing code whenever possible, in line
    with standard object-oriented techniques.
    Moreover, our approach allows for using existing 3D content,
    and for extensions based on Java technology, as indicated in A6. 

Q10. What applications do you have in mind?

A10. In ref [2] we discuss some case studies. Our approach might
    prove its worth in for example 'instructional VR', as testified
    by our Tai Chi demo (that is realized using STEP).
    Also the work of our candidate researcher indicates possible
    applications in agent-based task environments.



additional references:

[1] A. Eliens, Z. Huang, C. Visser,
    A platform for Embodied Conversational Agents
    based on Distributed Logic Programming
    http://www.cs.vu.nl/~eliens/research/media/title-platform.html

[2] Z. Huang, A. Eliens, C. Visser,
    STEP -- a scripting language for Embodied Agents
    http://www.cs.vu.nl/~eliens/research/media/title-step.html

[3] A. Eliens
    A (not so) gentle introduction to multimedia
    http://www.cs.vu.nl/~eliens/media

additional information:
    STEP -- http://wasp.cs.vu.nl/step
    XML-based STEP -- http://wasp.cs.vu.nl/step/xstep 

candidate researcher:
    Michiel Hildebrand (CKI, Utrecht)
    project page: http://www.cs.vu.nl/~mhildeb