A RANDOM PROCESS FOR OPTIMIZATION.
Abstract
A study was made of a method of optimizing a system by successive comparisons of the system performance for many randomly chosen combinations of the independent parameters. The study involved a Gaussian noise generator to randomly choose the values of the independent parameters. The system was examined for an optimum solution by repetitive operation of the analog computer for randomly chosen combinations of independent parameters. An error function was formed for each solution and the independent parameters for that solution were saved. New values for the parameters were randomly chosen from the Gaussian generator and a new solution was found. If the error for this solution was less than the previous error, the new values of the parameters would replace the old ones. In this manner, the proper values of the independent parameters were found for the maximum error allowable. This method allowed the complete range of interest to be scanned and functioned properly in the system chosen for investigation. The system chosen for investigation was a simplified helicopter feedback control system for a helicopter at hover. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 1969
- Accession Number
- AD0695654
Entities
People
- Victor Steve Vajo
Organizations
- United States Army Communications-Electronics Command