READAPT V&V

DARPATech PCA READAPT Exhibit

Lockheed Martin's Advanced Technology Laboratories (ATL) had an opportunity to host exhibits highlighting a number a DARPA Programs within the IPTO Booth at the DARPATech 2002 "Transforming Fantasy" Symposium. One of the four ATL exhibits highlighted the PCA READAPT Program. The PowerPoint presentation "DARPATech_PCA_Demo.ppt" prefaced the demo and provided an overview of the PCA Program as well as details of the demo given. (The PowerPoint file "Demo_Screenshots.ppt" captures some screenshots for the two demo displays.) Photos of the IPTO exhibit area are also shown below.

Demo Objective

• Problem
  • Current approaches to Verification & Validation (V&V) are not only cost prohibitive, but they cannot guarantee full state coverage for highly adaptable systems
    • PCA-based Systems (Morphable)
    • Cognitive Systems exhibiting human-like intelligence
      • Possibly non-deterministic - exacerbating the V&V problem
  
  
• Thesis
  • Run-Time Monitoring and Steering is a revolutionary solution for ensuring "safe" and properly executing behavior for PCA and Cognitive Systems based adaptable architectures
  
  
• Demonstrate:
  • An Efficient and Reliable means of capturing both system behaviors and system designs (model behaviors)
  • Ability to monitor complex, adaptable systems at run-time to behaviors captured during the design process
  • Ability to force a complex system into a properly behaving state in response to changing behaviors and behavior violations

Demo Approach

• "FlightViewer" with "Route Planner" Views for UAV waypoint navigation provides operational avionics context
• Route Planning Algorithm is implemented in CSIM on a model of a PCA based computing infrastructure
• CSIM PCA enabled Route Planner model will control FlightViewer Route Planning overhead display
• Route Planning behaviors are captured in MaCS/MEDL
• MaCS monitors PCA enabled Route Planner during "Auto Piloted" Flight Simulation
• A mission change (e.g. SAM pop-up) causes a route re-plan, requiring a PCA morph
• MaCS discovers a requirements violation and then forces the PCA enabled route planner into a "safe" morph

Route Planning Algorithm

• Onboard real-time route planning represents an adaptable algorithm (route planning algorithms could be considered cognitive)
  • Find the best path from the initial point to a target set of endpoints
  • Route is formed by a series of segments, each of which has an associated cost
  
  
• Well suited for PCA
  • Multiple possible resource adaptations
    • The number of directions at each grid point can be expanded (or reduced) to go to nearest neighbors, next nearest neighbors, etc.
    • The distance or scale between grid points can also be adjusted, as well as how far into the future the route can be planned subject to processor resource constraints (the area over which the replan takes place, the prime method of controlling constraints in the real-time environment).
  • The optimal route can be found by removing all processor constraints, extending the planning horizon to the target set, shrinking the grid size, increasing the number of directions at each point and finding the minimum cost path. This is a measure of performance for all approximations made in embedding the application in the polymorphic architectures.

	Demo Presentation Slides
Click on the picture above for an enlarged view of the slide

	IPTO Exhibit Booth
Click on the picture above for an enlarged view of the photos