Resource Analysis of Cognitive Process Flow Used to Achieve Autonomy
Abstract
The current challenge of autonomy is to achieve a reasonable scaling between task breadth and system resources. As processing resources are a limiting factor for autonomous operations in complex environments, the incorporation of new enabling low power processing technologies into autonomous systems is important to overcoming current limitations and keeping pace with peer adversaries. With the increasing variety of processing technologies, the number of design choices for implementing end-to-end cognitive processing flows multiplies and the impact of these design decisions on efficiency and effectiveness increases. The goal of this paper is to provide insights and guidance to system designers and program managers, not necessarily familiar with cognitive processing, regarding the resource/performance tradeoffs, and to provide guidance on the costs and benefits of different approaches to cognitive processing. This paper is organized into two parts: the first introduces an analytical framework within which the relationships between task complexity and system complexity can be formulated; in the second part, we introduce and analyze a canonical architecture called context switching cognitive processing architecture that exploits heterogeneous and run-time reconfigurable processing hardware to address the conflict between operating range (i.e., breadth), and efficiency through dynamic specialization of processing capabilities to the current task demands.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2016
- Accession Number
- AD1006453
Entities
People
- David Rosenbluth
Organizations
- Lockheed Martin