Dynamic Domain Architectures for Model Based Autonomy

1
Dynamic Domain Architectures for Model Based
Autonomy
New Ideas

• Executable Models with Explicit Constraints
• Multiple domain specific languages for physical
  constituents and software components
• Embedded in common base language, deductive
  framework
• Integration of models via fast runtime deductive
  constraint resolution
• Runtime compilation and optimization
• Models preserved for diagnosis in case of
  breakdown
• Three key services all model-driven
• Mode identification (diagnosis)
• Reactive planning
• Resource management

Impact
Schedule

• Toolkit for embedded SW framework developers
  helps to capture structure knowledge of the
  application domain.
• Embedded frameworks can be integrated with each
  other.
• Application programmers can implement and evolve
  application functionality at a high level, in
  concise, domain-appropriate, terms.
• Application users obtain
• more functionality in a more timely fashion
• they can rely on the correctness and performance
  of the application as it continues to evolve.

Reactive Executive for Model-Based Autonomy
milestones 1 design of next-gen Livingstone 2
implementation of next-gen Livingstone 3
testing and evaluation of next-generation
reactive executive. Developers' Toolkit for
Model-Based Autonomy milestones 1 design
meta-language describing DDA programs 2
prototype DDA-based Application Generator 3
test, evaluate DDA toolkit Integration,
Experimentation and Evaluation milestones 1
prototype testbed application 2 Application
specific toolkits for testbed 3Reactive
Executive for Model-Based Autonomy
MIT AI Lab B. Williams, H. Shrobe, R. Laddaga