A Software Control Framework for Learning Coordinated, Multi-Robot Strategies in Open Environments
Abstract
The UMass MARS team contributed technology and infrastructure for the control of adaptive sensory and motor processes. We developed a suite of techniques for capturing interesting process dynamics in specific run-time contexts in order to learn control decisions. Particular attention was paid to the ability of teams of robots to adapt dynamically to changes in environment and mission requirements. The UMass effort marries high-level process descriptions, discrete event analysis and model checking, learning and stochastic exploration, and a control theoretic substrate to accomplish these goals. Distributed control Technologies have been transferred to SPAWAR in San Diego, CA. "Whole-body" distributed manipulation controllers and finger gaiting code for autonomous manipulation tasks have been ported to NASA-JSC for use in the Robonaut program. With related funding under the DARPA DASADA program, contractors of the U.S. Army for the Rotorcraft Pilot's Associate (RPA) and * Theater High Altitude Area Defense (THAAD) programs are exploring the use of Containment Units (CUs) to build adaptable systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 22, 2003
- Accession Number
- ADA418131
Entities
People
- Roderic Grupen
Organizations
- University of Massachusetts Amherst