Use of Modal Analysis and Surrogate Solution Surfaces to Analyze and Assess Adaptive Autonomous Systems
Abstract
A method is proposed by which to analyze the logic of adaptive autonomous system (AS) models through diverse scenarios by treating them as black-box models and developing solution surfaces of these models by incrementing their input parameters over the whole parameter range. The different modes of the AS are expected to be recognizable on the solution surface. Gradient analysis can be used to identify the transition regions between modal zones. The use of surrogate solution surface calculations is proposed to ease the computational burden of the black-box analysis. The proposed method is expected to remain viable when the model input parameters are greater than two and visualization of the solution surfaces becomes difficult. Recalculating the solution surface in time as the AS model is put though a mission and analyzing changes in modal transition regions will indicate the adaptation, or learning, of the AS model as it performs its mission.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 01, 2018
- Accession Number
- AD1062481
Entities
People
- Patrick S. Debroux
Organizations
- United States Army Research Laboratory