(DURIP) UCSD AERODROME: ENABLING CLOSED-LOOP LEARNING ON RESOURCE-CONSTRAINED ROBOT TEAMS
Abstract
Our vision is that the new instrument will transform the Aerodrome into a state-of-the-art facility supporting onboard closed-loop data collection, indoor-outdoor transitions with complex obstacle configurations, wave simulation, wind disturbance generation, and autonomous recharging. This will create a quantum leap forward in the ability of the proposing PI team to test new autonomous system designs and algorithms, addressing fundamental challenges in distributed collaborative closed-loop learning for robot systems in real-world conditions. The new infrastructure will have a major impact on existing and future defense contracts, transforming core autonomy functions on mapping, navigation, collaborative learning and multi-agent coordination. It will enable research thrusts on online learning of contextual environment models and self-supervised learning of robot dynamics and associated closed-loop control policies, distributed across heterogeneous robot teams. The PI team combines researchers from 4 different engineering departments, ensuring interdisciplinarity and diversity in both research directions and technical approaches.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Aug 12, 2021
- Source ID
- FA95502010107
Entities
People
- Jorge Cortés
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, San Diego