Application of a Genetic Algorithm to the Optimization of a Missile Autopilot Controller for Performance Criteria with Non-Analytic Solutions
Abstract
Modern optimal control theory provides analytic solutions for a set of linear feedback design problems with linear quadratic performance criteria. Recent progress in the field of robust multivariable feedback design has incorporated additional constraints which have addressed the classical concerns with stability margins, system sensitivity and disturbance rejection. Despite these important advances, many practical design problems arise in which the desired system performance constraints cannot be accommodated by the available theoretic techniques. Genetic algorithms (GA's), on the other hand, offer a numerical search method which does not require a statement of the mathematical relationship between the performance criteria and the parameter update rule. The objective of this thesis is to demonstrate that GA's provide a method of optimizing control system problems with analytically intractable constraints. A linear missile airframe and actuator state space model is developed with linear feedback controller, and implemented in a discrete time simulation. A genetic algorithm is constructed to optimize the linear controller parameters, first with respect to a weighted linear quadratic performance index. Additional performance constraints are then imposed to meet rise time, peak actuator effort, and settling error specifications. Computer simulation results show that the genetic algorithm provides good convergence to near optimal controller designs for each successive combination of constraints.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1993
- Accession Number
- ADA343069
Entities
People
- Richard A. Hull
Organizations
- University of Central Florida