AFOSR YIP: Towards Preference-Aware Autonomy: Specification, Synthesis, and Interactive Planning
Abstract
Humans excel autonomy in their cognitive flexibility to make satisficing decisions in uncertain and dynamic environments. Preference plays a key role in determining what goals or requirements to satisfy when not all desired goals can be achieved. However, in contested and stochastic environments, it maybe infeasible for the human operator to decide what objectives to be compromised in a timely manner. It is critical to leverage autonomy to augment human’s decision-making capability by providing decision support and action plans. Such level of intelligence can only be achieved if the autonomous system is equipped with human-like preference-based reasoning for satisficing planning. This project aims at the major gap in theory and algorithms for preference-based decision-making in stochastic and dynamic environment, with complex mission objectives. Specifically, it focuses on three key questions: How to specify the human’s temporally evolving preferences and goals rigorously in formal logic? How to synthesize an autonomous system that provably satisfies the operator’s preferences in stochastic environments? How to make the system robust and adaptive when planning with initially incomplete knowledge about the preference and the uncertain environment?
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jan 21, 2022
- Source ID
- FA95502110085XX0
Entities
People
- Jie Fu
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- Worcester Polytechnic Institute