Presentation on theme: "1 Visualization of Uncertainty in the Common Operating Picture WorkGroup 4: Justin Hollands David Baar Thomas Porathe Liesa Lapinski Harvey Smallman Szymon."— Presentation transcript:
1 Visualization of Uncertainty in the Common Operating Picture WorkGroup 4: Justin Hollands David Baar Thomas Porathe Liesa Lapinski Harvey Smallman Szymon Stachniak Nada Ivanovic Carolyn MacGregor
2 Overview: Uncertainty Visualization Problem Decision-makers need to know what is not known as much as what is known Existing displays imply a ground truth that can mislead Objective Produce a COP that allows users to tailor their information needs on uncertainty to rapidly changing ROEs Approach - Develop a taxonomy of task-relevant uncertainties - Design and evaluate display concepts to query and visualize them Benefits Better informed, flexible decision making
3 Problem: Displays and “Ground Truth” Displays implicitly adopt uncertainty criteria Identity attributes: all certain Location & temporal: certain Identity attributes: category certain, platform/affiliation not Location & temporal: certain Risky More Conservative Criteria
4 Our Approach Develop and evaluate explicit visualisation concepts to Make uncertainty, underlying assumptions and display criteria explicit Query uncertainty levels of tracks Directly manipulate the consequences of choosing different uncertainty levels Support decision-making needs by tailoring uncertainty levels to extant ROEs.
5 Creating a Taxonomy of Uncertainty The different types of uncertainty inherent in visualizing the COP Uncertainty inherent in display technology »Display / data sampling (spatial, temporal) Uncertainty in the data displayed »Conflicting / missing attributes for track ID »Spatial and temporal sensor uncertainties Uncertainty in the perceptual response »Imprecision (e.g., with viewing angle). Weber fractions.
6 Existing representations of Uncertainty Limited array of existing methods for displaying uncertainty Track ID »Icon in icon (e.g. AADC threat coding) »Disintegrated attribute representations (CROs) Spatial uncertainty »Spatial envelopes (e.g. sonar, meterology displays) »Blobology (dispersion/distribution of troops across space – USMC) Temporal »???
7 Design & Control Issues What are appropriate designs for customizable displays for levels of uncertainty? How might querying of information be integrated with displays of target uncertainty? Should the customizable displays differ depending on the target? Global or track-specific queries? Integration of temporal information? Making the uncertainty information decision-relevant (response plan management)? What are the trade-offs between displaying uncertainty and display clutter?
8 Direct manipulation interface (70%) 100% 0% (10%) 0% 100% Certainty of ID Spatial Certainty
9 Direct manipulation interface (70%) 100% 0% (50%) 0% 100% Certainty of ID Spatial Certainty Time???
10 Direct manipulation interface (70%) 100% 0% (95%) M29 Certainty of ID 100% 0% 100% Spatial Certainty
11 Direct manipulation interface (95%) 100% 0% 100% Certainty of ID Spatial Certainty (95%)
12 Drill-down for assumptions etc (95%) 100% 0% 100% Certainty of ID Spatial Certainty (95%)
13 Integration with associated displays 0% 100% Assumptions 1 … 2… 3… 4... (updated) (95%) 100% 0% Certainty of ID Spatial Certainty (95%) Priorities 1… 2… 3… 4… (updated) Sensor Management/ Intel Management (updated) Control of Decision Rules of Engagement (updated)
14 Integration with response management (70%) 100% 0% (95%) M29 Certainty of ID 100% 0% 100% Spatial Certainty Shoot Query
15 Interpretation of Uncertainty How does viewpoint contribute to the operator’s assessment of uncertainty? Line of sight (occluded regions, exact location along the line of sight, cartesian location in 3D space) Distance estimation (relative vs actual Spatial resolution of display (aka blobology) What is the relationship between available time to decide and assessment of uncertainty as a useful decision- making variable in order to act? How should temporal representation be conveyed? What are the parameters of uncertainty that might be considered by an intelligent agent? Should uncertainty be displayed by an intelligent/automated agent?