8/7/2019 4. Bonneau - Software and Systems
1/19
Software and Systems14 March 2011
Dr. Robert J. Bonneau
Program Manager
AFOSR/RSL
Air Force Research Laboratory
AFOSR
Distribution A: Approved for public release; distribution is unlimited. 88ABW-2011-0781
8/7/2019 4. Bonneau - Software and Systems
2/19
2011 AFOSR SPRING REVIEW2311G PORTFOLIO OVERVIEW
NAME: Robert Bonneau
BRIEF DESCRIPTION OF PORTFOLIO:
Provide foundations for design and construction of
software-intensive systems operating in complex
environments
LIST SUB-AREAS IN PORTFOLIO:
Algorithms
Model-based synthesisLanguages
Analysis and verification
Human-machine interaction
8/7/2019 4. Bonneau - Software and Systems
3/19
Software & Systems Goals/Payoffs
Systems and Software Goals Enable quantifiable performance evaluation of critical
software systems
Manage software environments in order to preserve vitalmission functions
Comprehensively understand distributed effects in largesoftware infrastructures to predict global system failures
Payoffs
Enable evaluation of mission performance that is criticallydependent on software functions
Enable software and resource policies that guaranteeperformance of software in dynamic heterogeneousenvironments
8/7/2019 4. Bonneau - Software and Systems
4/19
Current Program Scope
Language-based assurance New programming languages or language constructs reduce
errors at run-time
Domain-specific languages enhance capabilities for code
generation Formal Analysis and Verification
Verification of system properties based on formal
specifications
Automated reasoning for detection of errors and model-basedsynthesis
Abstract models of systems and their interactions facilitate
automated generation of code
8/7/2019 4. Bonneau - Software and Systems
5/19
Current Program Scope
Advanced Algorithms for Real-Time and DistributedSystems
New algorithmic techniques provide theoretical basis for
more efficient and reliable operation in networked and
real-time environments Logic-based approaches for test selection Modeling
Human-Machine Interaction
Cognitive and mathematical approaches allow for
development of new models of human-machine
systems
8/7/2019 4. Bonneau - Software and Systems
6/19
Systems and SoftwareAgency Interaction
NITRD Coordinating Group High Confidence Systems and Software (HCSS)
NSF Software Design and Productivity
2010 Workshop on the Future of SoftwareEngineering Research
NASA V&V of Flight Critical Systems Ames Research Laboratory
Human Systems Integration Division Intelligent Systems Division
High Performance Computing Modernization Office Software and Many-Core
8/7/2019 4. Bonneau - Software and Systems
7/19
Systems and SoftwareOther funding agencies
Army Research OfficeSoftware investment mostly directed toward information
assurance
ONR
Software and Computing SystemsPrinciples for Correctness and Security Properties
Human Robot Interaction
Perception and Cognitive Control
NSF
- Software and Hardware Foundations
DARPA: Software Producibility
Part of Disruptive Manufacturing Technologies program
8/7/2019 4. Bonneau - Software and Systems
8/19
Systems and SoftwareProgram Trends
Model-based Synthesis
Formal Analysis and Verification
Advanced Algorithms for Real-Time and Distributed
Systems
Language-based approaches
Modeling Human-Machine Interaction
Agent-based approaches
8/7/2019 4. Bonneau - Software and Systems
9/19
Scalable Model CheckingC. Tinelli U Iowa, C. Barret, NYU
Approach: Formal verification suffers from state space explosion.
Compactly represent logical symbols in scalable nested satisfiability modulotheory (SMT)
Payoff: More automated more scalable verification to handle large
heterogeneous systems
Compact SMT Language Improved Lower DimensionalModel
8/7/2019 4. Bonneau - Software and Systems
10/19
Automated Model RevisionKulkarni, Mich State
Question : Is it possible to revise the model automatically such that it
satisfies the failed property while preserving the other properties?
Approach: Verification tends to use approaches that are fixed based on the notion
of pre-existing code and logical structures. In order to adapt to unanticipatedconditions it is necessary to be able to revise models if conditions change.
Payoff: In dynamic heterogeneous systems, it is necessary to update the
verification of the system as it evolves
Adaptive Verification Adaptive Verification
Space
8/7/2019 4. Bonneau - Software and Systems
11/19
Feedback in Software ArchitectureGrolku, Smolka, Stony Brook, Havelund, JPL
Approach: Many software systems are introduced into environments thathave uncertain conditions that result in unforeseen failures. How can
software systems be augmented to adapt to failures
Payoff: Systems such as those on networks or those subject to uncertain
physical environments can adapt to conditions.
Robust Architecture with Feedback Adaptive Software Feedback
Performance
8/7/2019 4. Bonneau - Software and Systems
12/19
Composable Software ServicesEdmund Durfee, UMichigan
Approach: Reliability and latency in software is difficult to compute in the
context of online management of real time software. Use composablesoftware design and optimization in scheduling to get the best reliability
and latency tradeoff.
Payoff: Allow time critical mission functions to be guaranteed under
dynamic heterogeneous constraints
Modular Composable Scheduling
Architecture
Resulting Latency vs.
Robust Performance Tradeoff
8/7/2019 4. Bonneau - Software and Systems
13/19
Runtime RepairS. Khurshid, UT Austin
Approach: A functional approach can be developed for real time softwareruntime repair using new paradigms for online verification
Payoff: Faults in software can be corrected in real time and tracked rigorously
Real Time Runtime
Software Repair ArchitectureResults in Corrections of
Multiple Faults
Mi i V ifi i
8/7/2019 4. Bonneau - Software and Systems
14/19
Mission VerificationElbaum, Dwyer U. Neb., Rosenblum, U. Col. London
Mission
Planning
Mission
Simulation
Mission
Deployment
Mission
ready
Program Analysis
Mission
Analysis
Approach: Develop a language to represent mission scenarios tied to
integrated distributed software architecture.
Payoff: Verify global mission properties as function of lower level
software constructs for quantifiable fault tolerance in achieving mission
objectives
ABCD
ABCD
ABD
ABD
AB
Mission AnalysisLanguage Architecture
Fault Tolerant
Mission Design
8/7/2019 4. Bonneau - Software and Systems
15/19
SPRUCEDrager/RI
Approach: Use parallel processing resources and network infrastructure asmeans of emulating and detecting system faults in new software deployment
Payoff: Deployment of new software tools has far fewer defects and more
detailed assessment of integrated system performance
8/7/2019 4. Bonneau - Software and Systems
16/19
VerificationHoman/RB
EnhancedAnalysis forVerification
Run-TimeArchitecture
Shift in VerificationParadigm
Shift in AssuranceParadigm
Reqmts
Design
Development
Test
V&V
Analysis
Analysis Analysis
Payoffs:
* Catch errors earlier
* Less costly analysis
* Improved coverage
* Run time techenforces safeoperating limits
Cost & Schedule
Confidence & Assurance
SystemsEngineering V
Approach: Global systems analysis must have a principled strategy forverification. Verification architectures must integrate with measurement.
Payoff: Less time will be spent debugging mission performance or forensically
discovering why failures occurred after mission deployment
8/7/2019 4. Bonneau - Software and Systems
17/19
Increased Scale/Integration via DSMLs Anchored in DEVS
(Douglass, 711th HPW/RH)
Navigator
Plans routes
from targets to
targets under
constraints
DEVS (discrete event
system specification) Formal rigor
Model reusability
Interoperability
S t d S ft
8/7/2019 4. Bonneau - Software and Systems
18/19
Systems and SoftwareAFRL Tech Directorate Interest/Coordination
Information Directorate Systems and Software Producibility
Multi-core Computing
Air Vehicles
Flight-critical systems and software
Mixed-criticality architectures
Human Effectiveness
Modeling of human-machine systems
Meta-information portrayal STTR
Robust Decision Making STT Large Scale Cognitive Modeling/C2WT
8/7/2019 4. Bonneau - Software and Systems
19/19
Transitions
Sztipanovits (Vanderbilt)
AFRL/RI using C2 Wind Tunnel to build composable environmentswithin which to design and conduct capability-based integrated C2experiments
Tinelli/Barrett (Iowa/NYU)
Rockwell-Collins interested in transitioning SMT-based verifierresearch into formal methods toolkits for avionics systems
Smolka/Havelund (Stony Brook/JPL)
JPL Mars Science Laboratoryusing rule-based specificationlanguage to ensure correct execution of software on next MarsRover
Durfee (Univ of Michigan)
Collaboration on SBIR with Intelligent Automation Inc., applyinghybrid scheduling techniques to large-scale human expert teamingproblems involving dozens of teams, hundreds of experts, andthousands of constraints.
Top Related