NASA Logo

NTRS

NTRS - NASA Technical Reports Server

Back to Results
Developing Humanoid Robots for Real-World EnvironmentsHumanoids are steadily improving in appearance and functionality demonstrated in controlled environments. To address the challenges of operation in the real-world, researchers have proposed the use of brain-inspired architectures for robot control, and the use of robot learning techniques that enable the robot to acquire and tune skills and behaviours. In the first part of the paper we introduce new concepts and results in these two areas. First, we present a cerebellum-inspired model that demonstrated efficiency in the sensory-motor control of anthropomorphic arms, and in gait control of dynamic walkers. Then, we present a set of new ideas related to robot learning, emphasizing the importance of developing teaching techniques that support learning. In the second part of the paper we propose the use in robotics of the iterative and incremental development methodologies, in the context of practical task-oriented applications. These methodologies promise to rapidly reach system-level integration, and to early identify system-level weaknesses to focus on. We apply this methodology in a task targeting the automated assembly of a modular structure using HOAP-2. We confirm this approach led to rapid development of a end-to-end capability, and offered guidance on which technologies to focus on for gradual improvement of a complete functional system. It is believed that providing Grand Challenge type milestones in practical task-oriented applications accelerates development. As a meaningful target in short-mid term we propose the 'IKEA Challenge', aimed at the demonstration of autonomous assembly of various pieces of furniture, from the box, following included written/drawn instructions.
Document ID
20150008649
Document Type
Conference Paper
External Source(s)
Authors
Stoica, Adrian
(Jet Propulsion Lab., California Inst. of Tech. Pasadena, CA, United States)
Kuhlman, Michael
(Rensselaer Polytechnic Inst. Troy, NY, United States)
Assad, Chris
(Jet Propulsion Lab., California Inst. of Tech. Pasadena, CA, United States)
Keymeulen, Didier
(Jet Propulsion Lab., California Inst. of Tech. Pasadena, CA, United States)
Date Acquired
May 20, 2015
Publication Date
December 1, 2008
Subject Category
Cybernetics, Artificial Intelligence And Robotics
Meeting Information
IEEE-RAS Humanoids Conference(Daejon)
Distribution Limits
Public
Copyright
Other
No Preview Available