Methodology for Comparison of Algorithms for Real-World Multi-objective Optimization Problems: Space Surveillance Network Design
Abstract
Space Situational Awareness (SSA) is an activity vital to protecting national and commercial satellites from damage or destruction due to collisions. Recent research has demonstrated a methodology using evolutionary algorithms (EAs) which is intended to develop near-optimal Space Surveillance Network (SSN) architectures in the sense of low cost, low latency, and high resolution. That research is extended here by (1) developing and applying a methodology to compare the performance of two or more algorithms against this problem, and (2) analyzing the effects of using reduced data sets in those searches. Computational experiments are presented in which the performance of five multi-objective search algorithms are compared to one another using four binary comparison methods, each quantifying the relationship between two solution sets in different ways. Relative rankings reveal strengths and weaknesses of evaluated algorithms empowering researchers to select the best algorithm for their specific needs. The use of reduced data sets is shown to be useful for producing relative rankings of algorithms that are representative of rankings produced using the full set.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 2019
- Accession Number
- AD1079683
Entities
People
- Troy B. Dontigney
Organizations
- Air Force Institute of Technology