Robust Fuzzy Autopilot Design Using Multi-objective Optimisation for a Highly Non-linear Missile


This thesis is focused on designing a robust nonlinear autopilot design for a highly nonlinear missile system in the presence of parametric uncertainties. First, Feedback Linearization is applied to the nominal missile model which produces an equivalent linear system. Applying linear control techniques, an outer loop is designed to drive the controlled variables to reach the required demand, hence the missile can follow a desired trajectory. Unfortunately the control law produced by the feedback linearization is not robust in the presence of uncertainties and hence in a real flight scenario will not be valid, and will exhibit nonlinear behavior for small changes in system parameters. Fuzzy logic trajectory control is then used in the outer loop to improve the robustness of the feedback linearization technique. An evolutionary genetic algorithm is then used to optimise the fuzzy control parameters. Multiple solutions (alternative fuzzy controllers) are obtained by using a Pareto based approach with non-dominated sorting. This has been combined with the reference point approach to incorporate preference information into the genetic algorithm to direct the search towards feasible areas which satisfy specified ranges on each objective. The design meets objectives defined on the closed loop performance: steady state error, rise time settling time and maximum percentage overshoot. From the multiple solutions the designer can choose the one which satisfies specified requirements. Fuzzy scheduled controllers are also used to control side-slip velocity for a large range of multiple demands. The design has been exercised for multi-model airframe dynamics at vertex points defined by 16 variables.