Explainable and Scalable Planning with Probabilistic Temporal Logic Specifications

From u-t-autonomous.info
Jump to: navigation, search

Funding source: NASA Early Stage Innovation
Technical point of contact: Jeremy Frank, NASA Ames
Period of activity: 2017-2020
NASA Grant Page
Presentation for the continuation review 2017

Overview of the Project

The objective of this project is to develop theory, algorithms and demonstrations for formal specifications and automated synthesis and learning of autonomy protocols for mission, resource and contingency management.

We partition the effort into three thrusts:

  • Compositional and hierarchical synthesis: Develop methods for automated synthesis from probabilistic temporal logic specifications.
  • Interpretable plans: Develop automata-learning methods to synthesize human-interpretable plans for practically infinite-state systems.
  • Explainable feedback: In cases in which there is no acceptable plan, develop explanations of the core reasons in terms understandable by humans.

All three thrusts address scalability explicitly through a series of algorithmic and architectural measures. The technical thrusts balance their general applicability and their impact and timeliness for human spaceflight operations.