Contributions to Arpa  CBD from NPAC - Principal Investigator 
Geoffrey Fox

Innovative Claims and Technical Description (A and G)
1)The WAVE adaptive networking environment will be integrated with an existing (but 
evolving) Java environment TANGOSIM supporting both collaboration and distributed 
simulation. This is implemented by generalizing the session manager of a Java 
collaborative system (such as the less powerful Habanero) to an event driven simulation 
engine. We will use the WAVE-TANGOSIM system to demonstrate an adaptive C2 
application based on one originally developed for Rome Laboratory in a static network 
environment. In this C2 application, videoconferencing and a Web technology based 
Geographical Information System (GIS) will adapt the video resolution (based on an 
existing H263 codec) and image (map overlay) compression level (existing wavelet 
system) to available compute and network performance. The C2 application will be 
entirely built using Web technologies (Java, VRML and C++ plug-ins) that we expect to 
be pervasively used in a broad range of devices from future Personal Digital Assistants to 
high-end graphics workstations. 
The C2 application will demonstrate how commanders and distributed warriors in the 
field with heterogeneous adaptive network interconnection, can send and receive 
information with adaptivity exposed at application and network layer. Services illustrated 
in this application include video and image compression, data fusion, world wide 
information systems and the planning. We will develop an enhanced script for the C2 
application to illustrate the key adaptive scaling features of this proposal.

2)Iterative distributed algorithms are the essential basis of future high confidence network 
and compute environments. However we know from other such iterative problems, that 
hierarchical multiresolution approaches offer substantial advantages over basic local 
iteration. These hierarchical methods combine iteration with cycling through the system 
viewed at different granularities with a coarse view initializing iteration at a finer 
resolution. This is successfully illustrated by the pyramid schemes in image processing 
and the multigrid approach to iterative partial differential solution. The methods allow 
information to travel through the full network much faster (logarithmic time). We will 
develop such an hierarchical approach MULTINET to the combined application and 
network management problem. This will use successful deterministic annealing methods 
developed for parallel computing data decomposition and other resource allocation and 
classification problems. 

3)The full adaptive C2 application with WAVE network and multiresolution resource 
management will be demonstrated on a heterogeneous network with an ATM backbone 
including NYNET(linking NPAC and Rome Laboratory) and the NSF vBNS of which 
NPAC is a node.

Deliverables (B)

1)Demonstration of a linked adaptive network and C2 application on a heterogeneous 
network with an ATM backbone and a range of associated clients and servers.

2)An integration of  WAVE with an advanced Web (Java) collaboration and distributed 
simulation engine TANGOSIM.

3)Hierarchical  adaptive dynamic resource management algorithms MULTINET applied 
to the linked adaptive network and application.

F)Milestones and Cost

Year 1: Standalone versions of adaptive hierarchical  resource management MULTINET 
and adaptive TANGOSIM C2 application.

Year 2: Prototype integration of TANGOSIM and MULTINET with WAVE using 
enhanced C2 script.

Year 3: Demonstration of complete adaptive C2 application on an adaptive network 
scaling number of information sources and clients.

Budget:                         Year 1   Year 2    Year 3
Task 1)Adaptive C2      $200K   $150K   $100K
Task 2)MULTINET       $150K  $150K    $100K
Task 3)Integration          $50K     $100K   $200K

Total                                $400K $400K    $400K

H)Comparison with Other Work

We believe that TANGOSIM environment and our prototype C2 application are the first 
to fully utilize Web technologies and so represent a model of future DoD applications 
built on the most rapidly evolving powerful COTS technologies - namely, the Web.
Hierarchical distributed resource management has been proven in several applications but 
we our proposed MULTINET when combined with WAVE represents a novel use of 
these ideas.
Adaptive wavelet and H263 have been compared to best available commercial offerings 
and are very competitive. As we wrote software, we can modify them for use in our 
adaptive application - this would impossible otherwise.

I Contracts

Two NPAC contracts are especially relevant:
Rome Laboratory: Collaborative Interactive Visualization, $2M from Sept 95 to Sept 97.
New York State: Living School Book and InfoMall, $750K per year ongoing. These 
funded much of  basic Web  technology including Web-based GIS.

J Previous Work
Incorporated in above