Notes from Fifth N/X Workshop, Aalborg, DK

Notes by Justin Hollands (DCIEM CA)

Wednesday, October 03, 2001

Erik Kjems, Director VR Media Lab

Reports directly to academic council—unusual
Independent of faculties, do a lot of independent work
Give faculty to work at one of the largest display centres in Europe

Bioinformatics, industrial design, data fusion, aviation dynamics, flight, tank simulation

Lisbeth Rasmussen DDRE DK, Zack Jacobson CAC, CA

Jacobson talks about N/X, IST Reference Model
Introduces Varga Varga as Chair of RTG 007

MartinTaylor, Martin Taylor Consulting, CA

Definition of visualisation
Evaluation for acquisition vs. eval for research vs. eval for iterative design
IST-05 Reference Model
Maps model to OODA loop
Six questions listed
Uses Santorini fresco picture as example of "visualisation system", goes through and asks the 6 questions wrt this fresco
Some questions answered better than others (first three better than latter three)
Translucency
Foregrounding/backgrounding—less important objects placed higher up, further back away from viewer
Fresco as interactive interface for "marine tasking order"
Click on elements of boat (captain, navigator) brings up relevant information e.g., for captain, basic mission statement; navigator-route map, etc.
Don’t go to places set back, go to welcoming people, up royal road, to theatrical area
Questions implied by questions ie.., Q1 becomes Q1a, Q1b, etc.
Eg.., what is user trying to achieve vs. Q1a who is the user?
In this case, captain, navigator, etc.

Q2 can user perceive there is progress
A2a, by scrolling whether ships have been assigned, departure and arrival locations,

ISSUE; How should the MTO Author be made aware if tasking has been completed. Put an yellow or red x when task is incomplete
What affects the user’s ability to perceive
A4.1 fresco interface is physically long, so related items may not obviously connect with each other
Solution: ensure that there are engines to test consistency among items such as route map and text, captain’s timing orders
Red and yellow x’s provide alerting capability (Q6)
Word "interactively" in definition of visualisation ties in with most of the questions

Engines, navigation, section (select relevant data), selection, various
7 Steps for evaluation
what user wants to achieve,
perceive,
choose,
ability to perceive,
ability to control,
two others

Zack Jacobson, CAC

Visual text mining for technology watch
Spot relations among elements of text in massive corpus—conceptual/meaningful relations
Identify clusters of similar items, and scant conceptual regions—concepts not represented in the data set
Search tasks: find the documents I need to read; find out where imp’t areas of confluence are, and areas that should have confluence but do not
Vision allows us to choose those 3D elements of the scene relevant to the task, e.g., driving
VITA visual interface for text analysis
Concepts identified by search engines top plane
Hits produced by search engines middle level
Bottom level is documents

Hits are small cylinders inside larger cylinder representing the document
Yellow spheres represent concepts
Artificial gravity moves private/sergeant away from other officer ranks
Second example shows that this approach not very scalable to large datasets

Autonomy visualizer (autonomy is a search engine)
Allows us to see how "V. Taylor" relates to "risk management" in Jacobson’s email
Color coding and size show previously visited pages "visual trail" (green cylinders are previously visited vs grey; size of smaller cylinder inside gets smaller as pages get older)
Nice clustering of nation and biological weapon (Iran-botulism; Iraq-sarin)

Asked six questions with respect to VITA

Use-Technology Watch
Treat taxonomy as query, and database as names of applications

Annette Kaster, FGAN

UkENDo
Support concept for creation and use of doctrines
Configuration of sensors and effectors on ship
System optimization of onboard systems

Parameter settings based on doctrines
Task: develop concept of use for work with doctrines
Development of demonstrator for the support concept

Demonstrator made of components
Situational data and doctrines feed into to main controller incl Inference engine
Controller incorporates working memory, and output sets operational parameters
Output produces displays to show values of operational parameters

need UI for inputting, editing scenario, own ship data
UI for inputting editing time tables and rules
UI for displaying operational parameters
UI for defining simulation time, starting simulation, and monitoring rules activity

Now with ref to IST-05 model
lists
Dataspace: op parameters, op settings, technical and tactical situations, and from doctrines
Engines
I/o devices
Visualising technical tactical situations
Rule structures/sequence
Inference process
Operator notification
Two types of user: navy officers, navy operators
Officers structure sort rules to create PDB/MDB
Trigger MDB by scenarios for tests/simulations
Operators need to monitor tactical situation k, and react to operator notification

Per Husmann Rasmussen, DDRE

Information Fusion
NATO Task Group on Information Fusion (TgonIF)
Want to develop a technology demonstrator for follow on work (TgonIFD)
Objective: JDL Data Fusion Level 2 (situation assessment) and Level 3 (Threat Assessment), in an ASC of a Combined/Joint HQ while in Peace Support Ops
May switch to anti-terrorism ops
Data fusion process model
Diagram
Sources, pre-processing, object refinement, situation refinement, threat refinement, leading to HCI
DBMS (support database, fusion database) feeds in at all levels
Level 4 Process Refinement

Tasks of TG on IF: Develop Conceptual Model of ASC, Develop and Evaluate Functional Model of the ASC, Identify Functions that could be Automated
Operational Preparation of the Battlespace (OPB)
Physical, socio-political and operational environment lead to OPB, lead to C2 command support
Physical environment breakdown
Social political environment: cabinet, media, CIMIC, host cultural considerations, allied and host political considerations
Operational environment G1, G2, etc.

ASC in Peace Support Ops
Military, disaster, socio-political, IS/COIN, and Crime Reports
Need to be classified, correlated, aggregated, and situation views created for presentation to the commander

ASC OODA Loop
G3 Cell (Commander) gives G@ Officerr Blue situation
G2 Orientat, Decide, then Act
Leads to Direction: Produce Collection Plan
Collect SANDAs, collate & evaluate,
Process (PCS), Process IPB, Disseminate to G3, G2, and other customers if good SA otherwise, seek additional information

Information Fusion
3 processes: classification, correlation, aggregation
class doees either ORBAT (organization of battle) independent or dependent classification of information
correlation reduces data set and enhances info content by fusing information contained in the reports
aggreagation procedss reduces detail by aggregating situation elements to a higher level entity
DFD data fusion demonstrator
e.g., Observe three tanks, determine unit type based on equipment type and category based on military knowledge (GEFC), show tank company symbol, then determine command level, and modify symbol accordingly

correlation e.g., wondering if tank company A moved from C or B, know that there is road to C not B, so C more probable

aggregation
situation three tank companies moving west
moving in pattern, if pattern matches template (e.g, staying about 1-2 km apart) presume they belong to same battalion
result new tank battalion symbol shown

scenario
IFOR multinational division (MND) IFOR into Banja Luke region of Bosnia-Herz

Themes:
Movementof forces back to barracks
Displayed population and ethnic resistance
Political structures, crime involvement,
Infrastructure needs,
Non Government Organization (NGO)

Conceptual model (level 0)
Divide insurgent cell into packages, divide packages among nations

Conclusion:
Conceptual model of MIC (multinational intelligence cell)
Operational demos for follow on project
Counterterrorism
HCI not Topic of TG on IF

Linas Bukauskas, Aalborg U

Observer Relative Data Extraction
Catalog of stars 50GB database
Visualization systems cannot show all data, also observer moving through space certain info shown, certain info not

Observer related data extraction
Show visible objects, show objects with specified visibility levels
Objects that will become (in) visible, objects that might be visible soon
Objects along a path

Create tree structure to access the data: tree orders objects according to visibility factor
Also a second storage access tructure B-Tree
Visibility factor VF is function of two parameters object, and observer
Visible objects function of DB, Obs, and rho
Show me all items that have a visibility factor greater than rho (criterion)
But this is insufficient because it ignores importance/size of items

Leads to need to cluster, hierarchy
VR is minmal bounding squre (MBS), brightness color could be incorporated
In MBS objects are visible
Have hierarhcial structure of MBRs and MBSs
Group clusters of MBRs in hierarchical sequence

MBRs internal notes
MBSs leafnodes
Can pack more objects into leaf 1kb nodes
Advantage for leafnode greatest with 3D objects
Three cases of queries: Perfect, Conservative, Optimistic
Ranges are as is, enlarged, reduced, respectively

point query (observer cursor position as input), vs. window query (region defined) treated in each case

related work: dynamic index structure for spatial searching R-Tree
X-Tree, SS, Sr, TPR trees
Kd tree, quad/Oct-Trees, etc.
Idea: Experimental work comparing human effectiveness finding targets using different algorithms
Idea: Way to transition between global and local views?
http://www.cs.auc.dk/~linb/ (Publications link) linb@cs.auc.dk

Eric Granum,

Demo in large circular display room, with stereo
Color data sets, scatterplots
Allows various Assignment of colors to conceptual dimensions
Can cycle through over time
Can see all as separate three way graphs in one space (lots of little 3 way graphs)
Idea: integration of mapping issues with visual momentum ideas—need way to transition between different coding approaches


Bill Wright, Visual Insights

CPOF: Observations regarding principles
Many visualizations over two years, lots of issues

Start with description of CPOF
Tailored visualizations: specific visualizations that align to user and to task domain
Functional, Inventive, and Process Objectives
Get rid of icons, bring in blobology
Membranes—synthesis of aggregation and details
Provide an at-a-glance comprehensibility of force attributes
Lots of experimentation
Process: See users, do design, test design
SMEs are active participants and provide requirements
Blobs should have see, shoot, and sense attributes
Dynamic force strneght varies with density power contact
Blobs relative to function

Composite blobs: show subentities
Have dimensions of strength
Representation treatments vs. behavior treatments
e.g., ring thickness vs. movement, density

do dynamic blos lead to greater understanding of important situation elements?
Do dynamic blobs help command discern patterns in the battlespace
Answer yes to both questions vs. control

Used simulation halt technique—importance of questions not assessed
Block Parties up to 1-6 Tactical Decision Games about 50

Force on Force Visualizations
Shows the CPOF displays
Bars used to represent attrition data
Incorporates brushing
Trying to show timeline data over geographic terrain, vectors fading over time
Circular blobs not working in urban terrain
Don’t get doubling of performance

Blob structure
Allow rich complex shapes for blobs that reveal details without overwhelming
Idea: How to map blob shape to task, measure effectiveness
Detection of direction of momentum

Five different conditions
Shoot spikes, use of yellow to indicate when engaged
Blob edge thicker direction of orientation
Vs. no shoot spikes, no directional blob edge

Best vs. complex (with shoot spikes and all icons) vs. simple (just significant icons, and just shoot spikes) vs. 2D traditional icons, flat map
Best worked best
20-30% improvement over traditional

details important to commander
Can we try to design with this in mind
Change blobs to suit what’s important at a given time
Got away from circles—circles were hiding important properties
Hiding relevant data
Outline/membrane approach—membrane being drawn around footprint of unit
Assumptions about data sources: radio frequency tags, digital compasses, tactical inference engines

Individual symbols shown when zoomed in; when zoomed out get the membrane

Some principles of graphic design
Design is purposeful, for specific audiences

www.mundidesign.com
use of color perspective, layout

observations ‡ principles
use a methology
engage users
allow two-way learning: users and technologists
be aware
application of principles tricky
don’t use color
the more specific the visulization the more effective it is
let the data speak
intuitive means no legend
placeholders wdont’ work
rez,‡ can’t substitute with fake data
generate blobs from entity data
don’t use blobs for fewer entities

General Discussion

Wright’s list—use of color discussion
Jacobson: I like color, want to be able to use it
M. Taylor: color coding issues—don’t use color hue to indicate quantity, but hue used to indicate a category (red forces vs. blue forces) is good
Color saturation can be used to indicate amount with some success, although straight luminance/greyscale probably works just as well
Wright noted that use color as last resort

Jan Terje Bjorke: Discussion of Wright’s displays ­how memorable they were vs. data shown in VR media lab
Hollands pointed out how the data types different

Wright made point about need for data to be realistic—accept no substitutes
V. Taylor: can’t always get real data

How to show uncertainty with blobs
Eric: Comparisons about visualizations
Augmented reality
Provides constraints, a formal language of a sort
Vs. complete freedom

Eric: Use of color—good for perceptual grouping
V. Taylor: need to know who users are

Try to be perfect: zack points out that can never do that
Wright says need principles, if you don’t get it perfect, people see the imperfection, distracted by them
Emphasis on medium over message

Eric: set reference for level of quality you want
M. Taylor: calibrating level of quality with attention
DREO story about radar; stopped Canada’s involvement in holographic radar
Fresco e.g., highlighting important information

V. Taylor: different display formats, devices, PDA vs. large display
Must transmit as little as possible in small devices
Wright: generals with laptop

Smestad: what is meant by "at a glance" visualization
Relates to Ward Page attention comment
Wright: time to comprehend reduced when "at a glance"
Bad presentation systems, given enough time, can be interpreted

Varga: Wrightl’s presentation on tailored visualization; what should the rest of us do
Wright says it’s diff between data viz vs. info viz.
Data viz has much less mapping between conceptual and perceptual dimensions

Jacobson: need incremental progress from one to another

Varga: don’t peak too early
Virtual reality as part of viz—using data and mapping it to time is one way to go from data to information visualisatio


Bukauskas: what about air force?
Wright: Army aviation (helos) move like tanks
Bukauskas: would you use color for blobs?
Wright: Blobs not relevant, they are abstractions for hundreds of entities

V. Taylor: # of user interfaces and what each shows
Wright: interesting issue

Da Silva Verissimo: why did you switch representational scheme for Vita
Jacobson: wasn’t clear how things clustered with large datasets
V. Taylor: liked old version b/c could open up page easily
V. Taylor: wants feedback
How to select in 3D, likes idea of putting crown of thorns in Eric’s display

Media lab @ Aalborg University

Demos and Tours

Birgit Holme Frederiksen

5000 visitors to centre
goal to establish centre of excellence for research and eduction, and projects with industry within visualisation and VR technology
passive stereo—one lens polarized horizontally; the other vertically
green vs. purple
Onyx 16 CPUs 2 Gb RAM, Polhemus fasttrack motion tracking
Lightwave matrix-hub—switches output among display units

CAVE floor is plexiglass 2.5 m x 2.5 m x 2.5 m
Back projection
Active stereo
1 observer with tracking and a few without


Thursday

Bill Wright

Brief discussion: me: Perfection vs. transparency—what was meant by perfection was really transparency

Topic: sonar tactical decision aid
Background, issues
Expected improvement in amount and quality of sonar data, and analytical models
ASW is a thinking war
Focus on one command level

Temperature in ocean varies in layers; temperature affects sonar
Sweet spots, shadows
Range on bearing—one dimensional in multidimensional world
Why not use isosurface—allows you to draw contours in a 3D volume
Got sketches from users

Came up with framework of linked views
Environment view (big), thumbnails, range summary (2d), key plan 2D map, acoustic volume analyzer (3D)
Acoustic volume analyzer: Range vs. bearing at a number of bearings makes 3D object, can see 2D versions on wall of space

Key plan—incorporates traditional symbols: make new work with old
Show areas that have been searched
Add or manipulate assets and asset tracks
"what if" planning tool

thumbnails
save the result set—good for a period of time
model runs taken at different times
eg. At 40, 100, 300 m

kept existing way of doing things—original 2D graph—activity at one bearing, different depths

what if playing with cutoff points on isovalues, isosurface
slider bars to change AN, FOM, POD—where you cutoff the values
ambient noise
FOM figure of merit
POD prob of detection
SV sound velocity

linked views—if changing depth in one view, other views should also be updated
if you play with your azimuth slice (bearing)
M. Taylor: how do you ensure that links are obvious
"make things live"
principle of dynamic query—ensure quick enough update rate
expose complexity of what operators dealing with to command level

Ross Walker/James Broad

Immersive Education Ltd.
Virtual Reality on Portable Devices
Large data on small amounts of silicon

Oxford U. Intel Education Initiative—why is software not doing a good job in education
Take enjoyment of games, bring into education environment
Games use large amount of illusion with minimum of hardware

Millions of polygons
Game called "Republic"
Populated city environment
Newspaper found on street gets read, bolts on metal handrails realistic—progressively more detailed info as zoom in, done with software
Integrated physics system
Set up role playing situations between characters and props
Dynamic lighting, shading

Discussion

Jacobson: do you have any potential applications (hammer without nails)—do you have any nails
Walker: not from this group

V. Taylor/Wright: Millions of polygons—why no problem?
Amount of redundancy in data set
Broad: Simplicity of architectural forms vs. natural objects
M. Taylor: pre-existing polygons vs. creating them on the fly
Walker: Parthenon demo—took 6 man-months to create
(in response to Wright’s question: how long does it take to build these models)
model of athena inside, generally model quite realistic
shadows within objects but not across

M. Taylor suggests that we use Smestad’s principles to discuss Wright’s visualizations
Smestad needed to show his guidelines
Hollands reminding M. Taylor/Wright about using M. Taylor’s guidelines vs. Wright’s visualizations
Sonar as domain
M. Taylor:explains his model
What does the user need to achieve: SA (Wright)
Wright: need task decomposition
Start with users: principal sonar/warfare officer, commander
PSO supervise sonar operators, involves equipment operation
M. Taylor: operators can be considered engines in the model
Hollands nature of data changes—commander getting/wants verbal declarative info., when operators data is probabilistic

To operators, engines operate on the raw sonar information
Engine algorithm understands current flows, temperature, salinity, etc.
Engines access historical database
Ray trace model
Output of model run compared to bathymetric data, etc.
M. Taylor: all these things in the dataspace
What info is sonar officer providing: probability or definitive
Active vs. passive sonar, deploy sonar or not, where to go, where to look
M makes analogy to chess player
Operator has to drive engines to look for patterns in oceans—shifts in arranged data

Bjorke: probablility concept of response tied to application;
M: app is reason for doing this
Wright what is transformation—is transformation between situations between where probability is relevant vs. not relevant
JT driving car 10 km/hr 10 m is long; but at 100 km/hr, 10 m is short
Different probabilities occur in each situation
M brakes on vs. not brakes on

Jan Terje Bjorke, Norwegian Defence Research Establishment, and Agricultural U. of Norway

Communication of geographical relations: Natural language and context dependent visualization
Perception of distance scaling
Perception of Topological relations

Contexts: user’s perceptual properties, properties of equipment, one other
Comm. Levels: syntactic, semantic level, pragmatic level

Map classes: reading maps, seeing maps, communicating maps
Bertin’s visual variables—changes in coding allow pattern to be seen
Processing maps ­one map per variable
Communicating map

Perceptual variables
Amount of information is limited by map users
In communicating map have grouped information into classes and by this reduced amount of information
Scaling symbols with changes in map scale

Algorithm: eliminate the most conflicting element until the value gets its max value
Max r value with condition optimal symbol size
Eliminate conflicting elements and comput an ptimal symbol size
Demonstrates application of Shannon’s information theory

Context determined by application area
Application of fuzzy logic
Use natural language sentences: e.g., distance from the airport to AOI is short
Apply membership functions 0-1 which describes probability statement is correct
Plots of membership functions for distance from airport, cheap hotel, etc.
Hollands wants all three—intersection operators from fuzzy logic—get a certain set of hotels
If strongly recommend assign Yager operator (e.g., 0.8), then only one hotel gets value greater than zero
Scatterplots are 2D maps

Iconic reps vs. linguistic sentences
Assignment of descriptors (small, large, very large) to graphical representations divided bar—an iconic representation of magnitude
Distance between points can be used as an icon—scale the distances with respect to frame zoom

Linguistic description of topological relations
Crisp vs. fuzzy relations: how much is B inside A
Eight possibilities: disjoing, meet, equal inside, covered by, contains, covers, overlaps
(2 x 2 table)
fuzzy interior vs. fuzzy boundary
developing icons to represent these concepts, and natural language word sequences
e.g., to low, medium, high degree B overlaps A, etc.

Conclusions: do not overload user with info
Need for map design based on soft computing—fuzzy logic
Natural language sentences based on evaluations of quantiatitve data at nominal level
General cartographic communication model—how to create communicating map
Excellent talk get this guy’s papers
Q&A: how to coalesce do not overload with give the users all the data
Wright: give them as much as possible with out overloading them; depends on user
Bjorke: Complexity comes from multiple views (over time)

Erik Kjems

VR for decision support in urban planning
Project at the VR Media Lab
Need to make model for visualiations so that ideas could be presented to town council
Model: reusing texture frees up memory (reference to earlier presentation
contourlines,
road-project
orthophoto
billboards (trees, hedges, lights, signs)—only use two polygons, always pointing against viewer
houses, photographed existing houses
other elements

primary req: as few polygons as possible, refresh rate at least 30 Hz
view should be as realistic as possible
consequences
use billboards, object splitting (split model into whole objects)—objects behind not in CPU
as smooth as possible surfaces
photographic low res textures
VML model, plugin from maxim

Presentation in virtual envionment concave display
First minutes of silence when show to town council
A few insecure questions
Then heavy discussion

Removed buildings that were placed badly, had no view
Added green areas
New infrastructure
Increased distance between buildings
Decided to allow only one story housing

Survey
Have you seen the environment before—how high are your expectations to the presentation
Did the presentation clarify issues in the project Yes

Did the presentation influence the acceptance of the project
People who sought that presentation already wanted it , so no real change
How important was mobility and iteractivity in the model? Important
Mono/stereo—mono was good for town model
Was presentation worth the money--$6-7K
Texture model made in 5 days; texturing takes time 18 students 6 weeks

Primarily the modelling work, secondary the large display system
Interaction was very important

Second round
Discussions got much more into detail and even more intensive

Center for 3D Geoinformation
$2.5 million project funded by EU
Large Scale 3D Landscape Information Model
140 km wide x 160 km high
1 m resolution
Laser, GIS< Aerial Photo, Databases
complete model of towns and rural areas around Aalborg (north Jutland)
can depicit temporal changes in structure of landscape last 10 years
Major Danish Mapand Geodata providers as partners
Lasserscanning used to detect landscape
Lasers make trees look like little mountains—laser coming straight down
Resolution of laser scanner
Can remove trees using algorithm
Dune formation
Data scanning provides terrain model
Photo superimposed
Can use horizontal laser scanning by driving past buildings

Forest
www.Blueberry3d.com
—swedish company
Allows tremendous resolution as you zoom in

Thematic data and 3D models—superimposing borders of fields over 3D model

Want to present over PDA

My question about design process affecting the kinds of decisions made, put greater distance between houses, ensure houses don’t block views, etc. may bias against costs
A: noncommittal, but yes the process does affect the decisions

Wright: Trees and houses separate data level? A Yes
Two source data sets height field and data overlay A multiple source, lots of databases
Wright transportation, systems populations, A yes sure, would like to extend into that area, but not right now
Wright could be useful for PDA
What kind of technical architecture—how to generate PDA material and deliver it

Discussion

Jacobson: Spend time on Smestad’s sheet.
Smestad’ places 13 guidelines on left
Visual summary Diagram on right
Uses the guidelines to assess one of Wright’s business diagrams floor/walls stock information
Diagram supports many of points
Hollands pointed out coding schemes issues
Dynamic motion makes interpretation of different colors more obvious—some columns hang down
Visual structure—how many units available at a time ­graph represents this

Captain Lars Abild Danish Army

DACCIS Presentation and Demonstration (Danish Army C2 Information System)
New Danish information system
Prototype system
Division breaking down into brigades
Danish army Almost entirely Danish division
NATO North East: Danish division with 12 PL div and 14 German div
Responsible for 710,000 km

5 projects within DACCIS
with each project are 4 phases
Operational User Group (OUG) set up especially for the DACCIS project, reporting directly to the AOC (Army Operational Command)
AMC Army Material Command, Maersk Data, OUG are three parties involved in project management
Brigade level—best guess organization 4 different HQs in brigade
not enough people for 24 hr period

DACCIS users should require less training than conventional system
Company educates instructors, instructors educate end users
Means end users don’t have to wait for next course at school

About half Danish Army are reserve

Developoment process include obtaining use cases = understanding tasks
Ie.., use cases = tasks
4 stages of workshop
goal: operational mid 2003

WS1 paper pencil,
WS2 look and feel
WS3 wizard of Oz ­get correct dialog, although may not be working properly

Hardware
AMC ­use COTS products ruggedized
Strong emphasis on ruggedization leads to abuse of equipment (gets dropped more often)
"never paint it green"—will lead to abuse

material has to function in an APC (army personnel carrier)
Phase 1 MSOffice, GIS, C2 planning and operational execution tools
Military message-handling system

Phase 2 much more functionality—specialized for G1…G6
Conceptual architecture: NT,MSSQL server
Need multiple servers: Application Server, Domain server, comms server

Interoperability difficult
DACCIS based on ATCCIS data model
Also AdatP3 for specific systems (air naval, heros, CWS, UAV)
For army, need information from neighbouring army units (via ATCCIS)

ATCCIS: working group create a data model, and replication model involving many NATO countries incl. US, UK, CA
Too much engineers, not enough army

Data Model is ATCCIS General Hub 4 (for phase 2)

Replication model ATCCIS Replication Model
Denmark and Norway only countries with small contracts

Currently at phase 1

Data model includes relations between data;
visualization is a national implementation
Wright’s question: Expert Users changing every 2 years; what do you do to figure it out
Answer: have to say OK, this was the best info we had, stick with original plan

Smestad: how many use cases?
A: about 200

Users want lots of map—most of DACCIS is map display
Scanned maps with scrolling and zoom
Move map with gesture
Select viewing vector by gesture (line of sight), get terrain profile, maintained on map afterwards
Area of sight also possible—what you can see from that location
Phase 2 will incorporate 3D button
Zooming by selection of rectangular area by gesture
Can select particular features to be plotted in the maps using check boxes
Select objective area by selecting circle symbol as in drawing program, again uses gesture to do this

Attack arrow—multi-headed possible based on user’s demand despite fact that violates NATO standard
Number of heads indicates phase/order of attack (this group attacks first, that second)
Select symbology from palette as in drawing tool

Task organization shown as hierarchy using Windows conventions (like win Explorer)
Selecting properties for particular units done with selection of unit, and modifying content of dialog box

Interactive whiteboard part of system
Problem with shadows/blocking
Developed special purpose stick for reaching far side of display
Different levels of commander have different access to plan—commanding officer can review/edit/approve plans of junior officer

Artillery position field of fire for different weapons done by selecting symbol from palette; angle range chosen and is adjustable to produce pie slice

Palette choices affected by G2-G4 check boxes
APP6alpha (APP6A) same as mil std 2525
Right click on symbol, select history, shows track
Old problem of different locations on different officer’s maps
Shared displays—all information now common
Outlook used for mail
Gesture recognition/keyboard at whiteboard
Symbols need to be scaled with zoom

Echelon
Combat effectiveness/combat
power—allowing a comparison of battle forces, our tank vs. their tank

Smestad: changing scale—computationally too difficult when scale is extreme
A: distance from map determines which symbols shown

V. Taylor: how much of phase 1 kept in phase 2
A; intention is to use it all; add new areas (e.g., helos, other Gs)
Programming will be redone; human interface same

Varga: use more saturated colors
A: yes plan to do that

Wright: deploy down to battalion level
A; yes

Wright what is purpose
A; Reduces # of staff meetings for planning

Unkown: What was granularity of Line of sight computation
A; 25 m

U: could you take it to Kosovo and use
A; Can take any map, geo reference it, and that is your source

Margaret Varga, QinetiQ, UK

Battle Damage Assessment Phase 1
Problems : what viz tools available
User involvement during process
Performance evaluation

Going to talk about BDA and evaluation usabilty problems
CAOC wide
Air tasking order data in spreadsheet
Get summary of planned mission in bar chart
Problem bar chart not scalable
Table viewer
Field search
Also have problems

Table links spreadsheet-color coded in terms of level

Visually query database
Select value that is required in each field
1D connections,
2D connections
Crown of Thorns 3D—show database linkages
2D using color
feature plot—table with color

other views shown
like map view

Discussion

V. Taylor: how do you get new technology to old spaces (people)
A: younger people more attuned to graphical representation
Gradual process of accommodation to new technologies

V. Taylor: research institutes looking at new technologies that may have application down the road
Difficulty occurs when get genuine user, they have their own tunnel vision
A: have better contact with military community; how much can tolerate

Wright: acceptance; reach further into user; get better measures
A: difficult because have to have controlled environment
Relate graphics performance to operational performance; say to user "can only get there if you can help us"
A: trying to do more trials, get deeper exposure to users; issues in getting unclassified databases
User Interest in database visualization—fusions of database
Wright yes it’s big time investment
A: need for geography

Jacobson: database viz nice way to start
Pilots in combat like to use systems they trained on
Make new systems part of training (links to V. Taylor’s comments about military colleges)
A: yes, have to persuade the young one; but old one still has authority

M. Taylor: high and low ranking officers
Bringing them in at all levels helpful politically
A: Do compromise so that user can accept

Da Silva Verissimo: should try to get new features in

Smestad: creating guidelines for non-specific domains is a challenge; can you explain
A: common solution is to have general set of tools within application; tries to do much
Smestad: relate to my guidelines
A:
Smestad: graphics were problems, not guidelines

M. Taylor: need generics, but need to make sure not overgeneralizing across applications
Can have guidelines at various levels: hierarchy
Guidelines can be made to be compatible
This is always true, in this situation do this, in that situation, do that
Always refining guidelines
M. Taylor says Smestad’s guidelines are of the general form

Smestad: My guidelines predict Varga’s results (are consistent with)
Failed to disprove

Smestad: Nature of hierarchy, relating to MIL-STD 1472D, guidelines need for hierarchy

V. Taylor: Easier to be generic the further you are from applications

Varga: NATO vs. TTCP: HF vs. operational

M. Taylor: Smestad’s guidelines fit for generic structure of engines
At lowest level that is not true very few things can do on a screen other than change a pixel
Various models slot in at different levels

M. Taylor: distinction between what we do to develop the the science, improve our understanding of how viz works; and what we do to improve an interface now, these are not opposing goals

V. Taylor: guidelines not absolutes
Varga: results against guidelines
V. Taylor: have constrained set of things you can handle

More talk about getting users interested in new features
Varga: will need more training

Hollands Need for multiple displays, map not only solution

Jacobson: note Wright wright’s attempt to have multiple views, seeing views sequentially
Varga: yes different displays important, different display modes suit different tasks

Wright: not happy with sonar , needs more iterations
Jacobson her problem is getting the iteration started

V. Taylor: considering the points of the last few days, what do people think of the IST-05 model?

Smestad says yes good general framework, not inconsistent with guidelines
Jacobson: works differently at different levels of hierarchy
people can be part of model
V. Taylor: how do we get out of this organization at the next levels
What happens if we have a peer relationship

M. Taylor: each one appears in the dataspace of the other

V. Taylor: something about hierarchy

M. Taylor: people who are affecting what you are trying to do are engines, whether hierarchical or not, engines can be people or silicon

V. Taylor: one of problems with creating chapter on engines, was this refinement on hierarchy

Smestad: not surprising that the model works; based on cybernetic principles initiated 19th century and refined by Norbert Wiener

V. Taylor: chief sonar operators, his model applied to sonar operator, his engines include what he views and hears, gets concept, his action is to alert someone that something is taking place
Understanding and acting fo sonar operator is output device for next level

Hollands scientific model is bad; as a practical model is good

M. Taylor: Six questions evaluator should ask
Iterative nature; use it recursively

Hollands Donald Norman’s control loop