Hover and forward flight of an Autonomous UAV using optimal linear control and gain scheduling
Authors
Bæk, Teis ; Hammelsvang, Mads ; Jørgensen, Thomas Bæk
Term
10. term
Publication year
2007
Abstract
Dette speciale udvikler lineære styringsstrategier, der skal muliggøre autonom flyvning for en AUV baseret på en omkonfigureret Bergen Industrial Twin-helikopter. Målet er IARC Level 1: at flyve 3 km så hurtigt som muligt. Udgangspunktet er en detaljeret ikke-lineær model af helikopteren, hvor lineære regulatorer designes ved at linearisere modellen. Der anvendes to metoder: klassisk single-input single-output (SISO) regulering, implementeret med lead- eller lag-kompensatorer, og optimal linear quadratic regulator (LQR) regulering. Et gain-scheduling-skema skifter eller blander mellem to regulatorer, efterhånden som driftsforholdene ændrer sig, ved at bruge en observatør til at estimere flyets tilstand og vælge passende gains. En overordnet supervisory-controller håndterer positionsstyring. Vi designer hover-kontrol med både SISO og LQR samt en fremadflyvnings-LQR-regulator til hurtig marchflyvning. LQR-hover- og LQR-fremadflyvningsregulatorerne samt den observatørbaserede gain-scheduler testes hver for sig mod tre Level 1-specifikationer for at sammenligne ydeevne. Den observatørbaserede gain-scheduling-tilgang var den mest tidseffektive af de testede metoder.
This thesis designs linear control strategies to enable autonomous flight for an AUV based on a reconfigured Bergen Industrial Twin helicopter. The target is IARC Level 1: flying 3 km as fast as possible. We start from a detailed nonlinear model of the helicopter and build linear controllers by linearizing that model. Two design methods are used: classic single-input single-output (SISO) control, implemented with lead or lag compensators, and optimal linear quadratic regulator (LQR) control. A gain-scheduling scheme switches or blends between two controllers as operating conditions change, using an observer to estimate the aircraft state and select suitable gains. A high-level supervisory controller handles position control. We design hover control using both SISO and LQR, and a forward-flight LQR controller for fast cruise. The LQR hover and forward-flight controllers, and the observer-based gain scheduler, are each tested against three Level 1 specifications to compare performance. The observer-based gain-scheduling approach was the most time-efficient of the tested methods.
[This abstract was generated with the help of AI]