Modular Affective Reasoning-Based Versatile Introspective Architecture (MARVIN)
Abstract
Report developed under SBIR Contract for Topic #SB072-009. In this Phase I study, we demonstrated that affect-inspired, self-aware mechanisms can significantly improve the performance of computational systems in cases of limited resources. We designed four affect-inspired adaptation mechanisms and argue that all of these mechanisms have the ability to positively impact the performance of various types of computational systems. We developed a MINIX-based implementation of an affective attention and resource management module. For evaluation and demonstration, we developed a UAV-based scenario that requires the UAV to respond rapidly to environmental events. We found that the affective system dramatically outperformed a comparable non-affective system. This was not, however, at the expense of non-critical operations as the critical response processes were able to dynamically reduce their resource usage in non-critical periods, enabling the other processes to be even more efficient than their non-affective counterparts.
Document Details
- Document Type
- Technical Report
- Publication Date
- Aug 14, 2008
- Accession Number
- ADA485732
Entities
People
- Mike Reposa
- Scott N. Reilly
- Sean Guarino