Learning Online Temporal Goal Networks
Abstract
We propose to develop a new extension to goal networks called a Temporal Goal Network (TGN), which will allow representing time constraints in a goal network. We then propose to learn the structure of TGNs for representing and solving complex constraints on Navy-relevant problems in which teams of agents cooperate to achieve a shared goal against one or more adversaries. We will formalize a theory of TGNs, create algorithms for learning TGNs, and evaluate TGNs on benchmarks that ensure TGNs scale to dozens of teams and hundreds of goals in dynamic scenarios. TGNs will substantially reduce the time to automatically learn hierarchical knowledge and allow faster transfer of past experience to new contexts. They will incorporate temporal reasoning and support goal sharing in order to track past and future commitments in teams.Learning Online Temporal Goal Networks
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Apr 29, 2020
- Source ID
- N000142012257
Entities
People
- Dana S. Nau
Organizations
- Office of Naval Research
- United States Navy
- University of Maryland