Can trained monkeys design flight controllers for hypersonic vehicles?

Abstract

The supersonic combustion ramjet is an as yet unproven propulsion system for hypersonic flight. Provided it can be developed into a practical vehicle, the ultimate success of sustained hypersonic flight will depend on configuring a robust and stable airframe-propulsion-control combination. To design the longitudinal flight controller for this inherently unstable vehicle we have applied a genetic algorithm, hence the trained monkeys metaphor in the title. Being a nondeterministic search method, there is no guarantee of generating a useful solution, yet given a little direction and enough time it is able to solve hard problems. The controller is built using fuzzy logic rules, directed at manipulating the vehicle's angle of attack through the actuation of symmetric elevators. A preset structure for the rules is used whereby the design task is to configure the control surface through selection of the rule consequents. To direct the search for a controller design, the genetic algorithm uses simulated flight responses to a range of initial conditions, without linearization of the vehicle model and dynamics. Results for the genetic algorithm designed controller show longitudinal stability and disturbance rejection.