Provably Correct High Confidence Human Robot Interaction
Abstract
The team proposes a one-year project as an extension of a current ONR MURI program. We will enhance and complete new theories and methods that we have developed for building provably correct decision making strategies for unmanned systems with humans in the loop. These include formal guarantees for the correctness of interactions between robots and humans, with robots incorporating models of human decision making, humans incorporating robot decision making models and so on. Our results have demonstrated human~s cognitive and motor abilities can be significantly expanded and enhanced by robot collaboration. They also reveal several remaining critical challenges.In this project, we will focus on complete what we refer to as rudiments of High Confidence Design of Human Robot Interaction, which mainly means correct-by-construction and fault tolerant. In particular, our proposed work will focus on three broad themes: 1. Adaptive learning algorithms for improvingcollaboration of human-robot teams. 2. 3-D geometric primitive estimation on point cloud with their use in robot control and interaction in virtual environments. 3. Flexible-as-human task representation to enable fault tolerant transfer learning for human robot collaboration. Finally, the three research themeswill be tightly integrated into a unified design methodology for high confidence human robot systems.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 07, 2019
- Source ID
- N000141912066
Entities
People
- S. Sastry
Organizations
- Office of Naval Research
- United States Navy
- University of California Regents