Perception-Based, Reactive, Temporal-Logic Planning For Autonomous Deck Operations
Technical point of contact: Dr. John Kinzer, Office of Naval Research
Period of activity: 2013-2016
Collaborators: Kostas Daniilidis
Overview of the Project
| The objective of the project is to develop methods and tools for the formal specification and synthesis of reactive autonomy protocols for operations in constrained and uncertain aircraft carrier deck environments under sensing/perception limitations and imperfections.
The harsh environment, tactical requirements, and human/autonomy interactions in deck operations impose unconventional challenges for protocol synthesis by limiting the quality of sensory inputs and availability of information used in the implementation of the protocols. We propose to tackle these challenges by formally characterizing the imperfections and limitations; incorporating them in the abstractions and specifications used in protocol synthesis; and developing new algorithms that are suitable for automatically synthesizing from the resulting abstractions and specifications.
The project has three segments which merge ideas from perception (Daniilidis) and formal methods (Topcu) and evolve in parallel with each other.
Using perception to extract semantics with uncertainty: Given sensory measurements, how can we feed a semantic description of static and moving components of the scene and the uncertainty in each expressed probabilistically to the reactive planners?
Integrated architecture for the perception and navigation modules for an autonomous vehicle.
A representative layout for aircraft carrier deck (see the source).
- Thrust I: Semantic mapping
- Thrust II: Segmentation and semantic mapping of moving objects
Reactive synthesis with limited sensing & imperfect perception: How can we incorporate the uncertainty models developed in the perception segment into reactive synthesis problems with limitations in information availability and quality? How can we exploit the underlying system-theoretic interpretations to alleviate the resulting computational complexity?
- Thrust III: Synthesis & temporal-logic games with partial information
- Thrust IV: Synthesis in stochastic domains with imperfect information
- Thrust V: Coping with the computational complexity
Case study: How can we develop a case study that features the characteristics of the uncertainties and constraints of the operations on the deck environment and serve as a demonstration for the advances in the other two segments of this project?