Fixed-wing UAV Modeling and State Estimation
Author
Larsen, Hans Kresten Nørgaard
Term
4. term
Education
Publication year
2024
Submitted on
2024-05-31
Pages
55
Abstract
This project investigates two ways to model the flight behavior of a small fixed-wing drone (mini-UAV). The first builds a physics-based model using only known geometry and mass properties. The second learns a model from flight data using SIDPAC, a system identification toolbox for aircraft. In addition, the work presents a proof-of-concept method to estimate airspeed from the electric motor’s field-oriented control (FOC) signals combined with standard propeller equations and experimental measurements. Both modeling approaches require substantial setup effort but are useful once implemented. The geometry-based model depends on parameters that are difficult to obtain in practice: the products of inertia (which describe how mass is distributed), the maximum angle of attack, and the non-dimensional lift coefficient at that angle. The data-driven system identification works best when carefully designed inputs are applied directly to the actuators (e.g., control surfaces or motor) to avoid correlated signals and ensure sufficient excitation of the system. An airspeed estimate derived from FOC variables was obtained, but its accuracy could not be validated within this project.
Dette projekt undersøger to måder at modellere flygeadfærden for en lille fastvinget drone (mini-UAV). Den første opbygger en fysikbaseret model udelukkende ud fra kendt geometri og masseegenskaber. Den anden lærer en model af flyvedata ved hjælp af SIDPAC, et systemidentifikationsværktøj til fly. Derudover præsenteres en proof-of-concept metode til at estimere lufthastighed ud fra den elektriske motors feltorienterede styringssignaler (FOC) kombineret med standardligninger for propeller og eksperimentelle målinger. Begge modelleringsmetoder kræver et betydeligt opstartsarbejde, men er nyttige, når de først er implementeret. Den geometribaserede model er afhængig af parametre, der er svære at skaffe i praksis: inertiprodukter (som beskriver, hvordan massen er fordelt), maksimal angrebsvinkel og den dimensionsløse løftekoefficient ved denne vinkel. Den datadrevne systemidentifikation fungerer bedst, når nøje designede input gives direkte til aktuatorerne (fx styreflader eller motor), for at undgå korrelerede signaler og sikre tilstrækkelig excitation af systemet. En lufthastighedsestimation baseret på FOC-variable blev udledt, men dens nøjagtighed kunne ikke valideres inden for projektet.
[This apstract has been rewritten with the help of AI based on the project's original abstract]
Keywords
UAV ; UAS ; GNSS-denied ; FOC ; Airspeed estimate ; SIDPAC ; Systemidentification ; 6DOF ; parametric model ; non-dimensional coefficient ; Fixed-wing ; drone