A month-long quasi-experiment was conducted using a distributed team responsible for modeling, simulation, and analysis. Six experiments of three different time durations (short, medium, and long) were performed. The primary goal was to discover if synchronous collaboration capability through a particular application improved the ability of the team to form a common mental model of the analysis problem(s) and solution(s). The results indicated that such collaboration capability did improve the formation of common mental models, both in terms of time and quality (i.e., depth of understanding), and that the improvement did not vary by time duration. In addition, common mental models were generally formed by interaction around a shared graphical image, the progress of collaboration was not linear but episodic, and tasks that required drawing and conversing at the same time were difficult to do.

As collaboration in virtual environments becomes more object-focused and closely coupled, the frequency of conflicts in accessing shared objects can increase. In addition, two kinds of concurrency control "surprises" become more disruptive to the collaboration. Undo surprises can occur when a previously visible change is undone because of an access conflict. Intention surprises can happen when a concurrent action by a remote session changes the structure of a shared object at the same perceived time as a local access of that object, such that the local user might not get what they expect because they have not had time to visually process the change. A hierarchy of three concurrency control mechanisms is presented in descending order of collaborative surprises, which allows the concurrency scheme to be tailored to the tolerance for such surprises. One mechanism is semioptimistic; the other two are pessimistic. Designed for peer-to-peer virtual environments in which several threads have access to the shared scene graph, these algorithms are straightforward and relatively simple. They can be implemented using C/Cϩϩ and Java, under Windows and Unix, on both desktop and immersive systems. In a series of usability experiments, the average performance of the most conservative concurrency control mechanism on a local LAN was found to be quite acceptable.

No abstract

Virtual reality technology is increasingly being applied to globally distributed teams engaged in collaborative product design. Observations of product design teams have suggested four distinct patterns of collaboration-complementary, competitive, peer-to-peer, and leader-follower. Another insight from observation is that collaboration consists of fluid transitions between these patterns in the accomplishment of the design task, driven by a flexible process of subgrouping and regrouping which reflects the structure and progress of the task. Yet most collaborative virtual environment systems support only one pattern of collaboration-peer-to-peer-and those that do explicitly support multiple patterns or roles do not allow fluid transitions between them in the context of the same task. In addition, no explicit support is provided to allow subgroups to be formed and dissolved. A collaborative virtual environment that supports multiple collaboration patterns and fluid transitions was developed using the Shared Simple Virtual Environment (SSVE) application framework. A novel user interface widget, the collaboration tree, was created to drive the subgrouping and regrouping process. Group experiments were performed to test the operating hypothesis that support for group collaboration patterns led to higher performance. The result was that the operating hypothesis was confirmed; however, the conceptual approach to problem solving, suggested by the presence of support for collaboration patterns, may have been more significant than the actual mechanism provided.

i%e Simulation IntranetLProductDatabase Operator (SI.DO) project has developed Web-based distributed object architecturefor highpe~ormance scientlj7c simulation. A Web-basedJava inte~ace guihs &signers through the akrign andaiudysis cycle via solid and analytical moa%ling,meshing, jinite element simulation, andvariousforms of visualization. l%e SILPDOarchitecture has evolved insteps towmu?rsati&ying Sandia's Iong-term goal of providing an end-to-end set of services for high jidelityjldlphysics simulations in a high-peq+ormance,distributed and distance computing environment. Zhispaper &scribes the continuing evolution of the architecture toprovih high-peq40nnance visualization services. Extensions to the SIPDO architecture allow web access to w"sualizationtools that run on MP systems. l%is architecture makes these tools more easily accessible by providing web-based interfaces and by shielding the user~om the a%taiisof these computing environments. Tile &m"gnis a multi-tier architecture, where the Java-based GU7 tier runs on a web browser andprovides image display and control jmctions. l%e competition tier runs on A@ machines. The micililetiersprovide custom communication with MP machines, remotefile selection, remote Iaunching of services, load balancing, and machine selection. 7he architecture allows mialileware of various types (CORBA, COM, R.&U, sockets, etc) to connect the tiers using ahpters. The system allows for aaliing and removing of tiers depending upon the situm"on. Testing of constantly developing visualization tools can be done in an environment where there are only twotiers which both run on desktop machines. Z%isahwsft esting turnaround and &es not use compute cycles on high-pe~onnance machines. Once the code and inte~aces are tested they are moved to highperfiormance machines, and new tiers are aalied to handle the problems of using these machines. Uniform inte~aces are used throughout the tiers to allow this~exibility. &pen.ments test the appropriate level of interface: either a large set of specl~cfinction cd[s or a small set of generic function calls. I%isarchitecture is basedon the goals and constrm%tsof our environment: huge &ta volumes (ttit cannot be easily moved), use of multiple micililewareprotocols, MPpiafonn portability, rapid development of the visualization took, distributed resource management (of MP resources), and the use of existing visualization tools. Thisworkis supported by SandiaNational Laboratories, a multi-program laboratory operated by SandiaCorporation (a Loakheed Martin r.mm-md fnrtheTT.itwl .Stntw 11-mmtm.-nt of17nPT.N 11.Aw rmttm.tIW-A