Motor-response learning at a process control panel by an autonomous robot

Conference · Fri Jan 01 04:00:00 EST 1988

OSTI ID:5127828

Spelt, P F; de Saussure, G; Lyness, E; Pin, F G; Weisbin, C R

The Center for Engineering Systems Advanced Research (CESAR) was founded at Oak Ridge National Laboratory (ORNL) by the Department of Energy's Office of Energy Research/Division of Engineering and Geoscience (DOE-OER/DEG) to conduct basic research in the area of intelligent machines. Therefore, researchers at the CESAR Laboratory are engaged in a variety of research activities in the field of machine learning. In this paper, we describe our approach to a class of machine learning which involves motor response acquisition using feedback from trial-and-error learning. Our formulation is being experimentally validated using an autonomous robot, learning tasks of control panel monitoring and manipulation for effect process control. The CLIPS Expert System and the associated knowledge base used by the robot in the learning process, which reside in a hypercube computer aboard the robot, are described in detail. Benchmark testing of the learning process on a robot/control panel simulation system consisting of two intercommunicating computers is presented, along with results of sample problems used to train and test the expert system. These data illustrate machine learning and the resulting performance improvement in the robot for problems similar to, but not identical with, those on which the robot was trained. Conclusions are drawn concerning the learning problems, and implications for future work on machine learning for autonomous robots are discussed. 16 refs., 4 figs., 1 tab.

• Research Organization: Oak Ridge National Lab., TN (USA)
• DOE Contract Number: AC05-84OR21400
• OSTI ID: 5127828
• Report Number(s): CONF-880657-1; ON: DE88005055
• Country of Publication: United States
• Language: English

Related Subjects

42 ENGINEERING
420200* -- Engineering-- Facilities
Equipment
& Techniques
99 GENERAL AND MISCELLANEOUS
990210 -- Supercomputers-- (1987-1989)
ARTIFICIAL INTELLIGENCE
COMPUTERS
EXPERT SYSTEMS
HYPERCUBE COMPUTERS
LEARNING
PERFORMANCE TESTING
ROBOTS
TESTING