Platformsudvikling, sensormodellering og estimation af en Draganflyer X-Pro
Translated title
Platformdevelopment, sensormodelling and estimation of a Draganflyer X-Pro
Authors
Haukrogh, Jesper ; Binderup, Ole
Term
10. term
Publication year
2007
Pages
243
Abstract
Afhandlingen omhandler platformudvikling, sensormodellering og tilstandsestimering for quadrotor-helikopteren Draganflyer X-Pro med henblik på at sammenligne to estimatorer af forskellig kompleksitet og at implementere den simple estimator sammen med en simpel controller. Platformen blev instrumenteret med sensorer og software til autonom flyvning, inklusive detektion af sensorfejl og mulighed for manuel styring med en stabiliserende hjælpekontroller. Alle monterede sensorer blev modelleret (bl.a. IMU og GPS) for at karakterisere deres målinger; denne viden blev anvendt til præfiltrering, tilstandsudvidelse og parameterinitialisering for både et Steady-State Kalman-filter og et Unscented Kalman-filter. Begge filtre blev designet og afprøvet på data fra en testflyvning og tilpasset til at håndtere forskellige sampelfrekvenser på tværs af sensorer samt de særlige egenskaber og begrænsninger ved quaternion-baseret attituderepræsentation. Implementeringen af den simple estimator er endnu ikke flugttestet på platformen. Afhandlingen afslutter med forslag til forbedringer, herunder videre arbejde med forbedret attitudebestemmelse.
This thesis addresses platform development, sensor modeling, and state estimation for the Draganflyer X-Pro quadrotor, aiming to compare two estimators of different complexity and to implement the simpler estimator alongside a simple controller. The platform was instrumented with sensors and software to enable autonomous flight, including detection of sensor faults and an option for human-pilot control supported by a stabilizing controller. All mounted sensors were modeled (including IMU and GPS) to characterize their measurements; this knowledge informed pre-filtering, state augmentation, and parameter initialization for both a Steady-State Kalman Filter and an Unscented Kalman Filter. Both filters were designed and evaluated on data from a test flight and were adapted to accommodate sensor-specific sample rates and the properties and constraints of a quaternion-based attitude representation. The simple estimator has not yet been flight-tested on the platform. The thesis concludes with proposed improvements, notably further work to enhance attitude determination.
[This summary has been generated with the help of AI directly from the project (PDF)]
Documents