Design of a Pilot-activated Recovery System Using Genetic Search Methods

Control design tasks often require the use of trialand-error search methods to obtain a satisfactory solution. Depending upon the nature and the number of “tuning” parameters or functions, the search process can be very discontinuous and nonconvex. The genetic search methods are a recently developed family of techniques for optimization which offer certain advantages over other techniques. These include greater freedom in defining cost functions and constraints, and the ability to automate the design process. Most notably, though, is the ability to construct new control laws and the potential to generate non-intuitive solutions as well. This paper demonstrates the application of genetic search methods to design a Pilot-Activated Recovery System (PARS) for a modern fighter aircraft. The PARS is a guidance law that transfers the aircraft from any initial attitude to a wingslevel, nose-up, recovered flight condition. This system is useful in cases of pilot disorientation. A 6 degree-offreedom nonlinear model of a modern, highperformance aircraft is used for design. The genetic search seeks to produce nonlinear feedback functions to meet the specified goals and constraints. This intricate problem highlights some of the advantages of this emerging search technique.