Comparing Adaptive Control of Thought-Rational (ACT-R) Baseline Activation Terms for Implementation in the Symbolic and Subsymbolic Robotic Intelligence Control System (SS-RICS)
Abstract
The U.S. Army Research Laboratory is developing the Symbolic and Subsymbolic Robotic Intelligence Control System (SS-RICS) as a robotic control system based on human cognition. It has, as part of its intelligence, a production system similar to that of Adaptive Control of Thought Rational (ACT-R). One feature of the ACT-R system is that chunks of information about the surrounding environment have recorded activation values, which increase when the system turns its attention toward them and decrease in the absence of attention. This feature can be beneficial in keeping recently used information readily available to attention. It can also be detrimental when a feedback loop develops, with a chunk being selected for attention because it has the highest activation and then increasing its activation value as a result. This report explores the behavior of the baseline activation equation with different parameter settings, and over different patterns of selection, in order to provide SS-RICS researchers with information for selecting appropriate parameters. It also makes recommendations as to how the equations might be modified to eliminate such feedback loops.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 2014
- Accession Number
- ADA600879
Entities
People
- Craig T. Lennon
Organizations
- United States Army Research Laboratory