Excitation and Estimation for Adaptive Control of a Supersonic Missile

Sammanfattning

Most missile systems today use model based controllers. The performance of such controllers will be determined by the accuracy of the model compared to the true system. If production is scaled up or the production cost of the missiles produced needs to be decreased, there will be a larger deviation between individual missiles. One way to mitigate the problem of model uncertainties is to use an adaptive controller where the control law changes in order to improve performance mid flight. In this thesis, such methods will be explored in combination with methods that force adaptation to be rapid and accurate. For the missile to gather useful information to be used during adaptation of the control law it will have to manoeuvre somehow comparing its motion to the predicted motion. Given that the goal of a missile launch is to strike a target the performance in the later stage is more important than the initial performance, therefore it is important that adaptation happens early. Thus, initial excitation through Optimal Experiment Design is considered. Another method of mitigating the risk of poor performance late in the mission stage is to enforce persistence of excitation throughout the entire flight and constantly adapt to changes in parameters. The adaptive controllers in this thesis are indirect, meaning that aerodynamic coefficients are estimated and then used in a new model based control synthesis. Two estimation methods are considered, Recursive Least Squares and the Prediction Error Method. The considered controllers are: an Adaptive Gain Scheduled Linear Quadratic Controller, Adaptive Model Predictive Controller and a Persistently Exciting Adaptive Model Predictive Controller. The results show that performance is improved with the adaptive strategies. What is also found is that the robustness of the adaptive schemes is higher than the pre-implemented Gain Scheduled Linear Quadratic Controller via Monte Carlo simulations.