Stabilization of a rotary wing unmanned aerial vehicle with an unknown suspended payload

Erasmus, A. P. (2020-03)

Thesis (MEng)--Stellenbosch University, 2020.

Thesis

ENGLISH ABSTRACT: This thesis addresses the problem of stabilizing a quadrotor with an unknown suspended payload able to swing in one axis. The payload parameters, such as its mass and cable length, are unknown and its swing angle is not available for measurement. The suspended payload significantly alters the flight dynamics of the vehicle as it induces oscillations into the system. This project attempts to design, simulate and practically demonstrate a control strategy to damp these oscillations and maintain stable flight. Two control strategies were explored, namely a Linear Quadratic Gaussian (LQG) control approach and a robust Model Reference Adaptive Control (MRAC) approach. Both of these control strategies were explored in simulation, of which one was chosen to implement on a practical vehicle. The LQG control approach estimates the payload parameters and its swing angle. Optimal full-state feedback control is implemented to simultaneously control the vehicle and the payload swing angle. The MRAC scheme adapts its controller to change the closed-loop dynamics of the system to that of a predefined reference model. Both strategies were extended to perform well in the presence of external disturbances and sensor noise, of which practical systems su↵er. They proved to sufficiently damp the oscillations caused by the payload, but the MRAC scheme has more advantages such that it consists of fewer components and provides consistent performance with di↵erent payloads. The MRAC scheme was chosen to implement on a practical vehicle. A quadrotor vehicle was built to practically demonstrate the e↵ectiveness of the proposed MRAC scheme. The avionics of the vehicle consists of a Pixhawk flight controller running the PX4 flight control stack. The control gains of PX4 were updated according to the custom-built quadrotor and the MRAC scheme was implemented in the flight stack. Software-in-the-Loop (SIL) and Hardware-in-the-Loop (HIL) simulations were performed with the Gazebo simulator to ensure that the implemented MRAC scheme worked as expected in the PX4 environment. After the success of the simulations, practical flight tests were performed to demonstrate the e↵ectiveness of the MRAC scheme. Three flight tests, each with a di↵erent payload, were performed. The MRAC algorithm successfully damped the payload oscillations in each flight and even outperformed a PID controller, tuned for the specific payload, in terms of reference following. MRAC adapted its control parameters to accommodate the specific attached payload in each flight. The flight tests were successful and the algorithm proved to damp the oscillations caused by the payload while maintaining stable flight.

AFRIKAANSE OPSOMMING: Die fokus van hierdie tesis is die stabilisering van ’n vierrotor onbemande vliegtuig met ’n onbekende swaai vrag wat in een as kan swaai. Die loonvragparameters, soos die massa en kabellengte, is onbekend en die swaaihoek is nie beskikbaar om te meet nie. Die swaai vrag verander die vlugdinamika van die voertuig aansienlik, aangesien dit ossillasies in die stelsel veroorsaak. Hierdie projek poog om ’n beheerstrategie te ontwerp, te simuleer en prakties uit te voer om hierdie ossillasies te demp en stabiele vlug te handhaaf. Twee beheerstrategieë word in hierdie tesis ondersoek. Dit is ’n Lineêre Kwadratiese Gaussiese (LKG) benadering en ’n robuuste Model Verwysing Aanpasbare Beheer (MVAB) benadering. Albei beheerstrategieë word in simulasie ondersoek, waarvan een gekies word om op ’n praktiese voertuig te implementeer. Die LKG benadering bereken die loonvragparameters om die swaaihoek af te skat en maak gebruik van optimale terugvoerbeheer om beide die voertuig en die swaaihoek te beheer. Die MVAB skema pas sy beheerder aan om die geslote-lus dinamika van die stelsel te verander na dié van ’n vooraf gedefinieerde verwysingsmodel. Albei strategieë is uitgebrei om goed te presteer in die teenwoordigheid van eksterne versteurings en sensor ruis. Albei strategieë het die ossillasies wat deur die loonvrag veroorsaak word voldoende gedemp, maar die MVAB skema het meer voordele soos di´e dat dit minder komponente het en dat dit konsekwente resultate lewer met verskillende loonvragte. Die MVAB skema word daarom gekies om op ’n praktiese voertuig te implementeer. ’n Vierrotor voertuig was gebou om te gebruik vir praktiese vlugtoetse. Die avionika van die voertuig bestaan uit ’n Pixhawk vlugbeheerder wat die PX4 kode hardloop. Die beheer aanwinste van die PX4 beheerargitektuur was aangepas vir die vierroter en die MVAB skema was geïmplementeer in PX4. Beide sagteware-in-die-lus en hardware-in-die-lus simulasies was uitgevoer met die Gazebo simulator om te verseker dat die geïmplementeerde MVAB skema werk in die PX4 omgewing. Na die sukses van die simulasies, was praktiese vlugtoetse uitgevoer om die MVAB skema prakties te demonstreer. Drie vlugtoetse, elk met ’n ander loonvrag, was uitgevoer. Die MVAB algoritme het die loonvrag ossillasies in elke vlug suksesvol gedemp. Dit het selfs beter gevaar as ’n PID beheerder, ontwerp vir die spesifieke loonvrag, in terme van die volg van die verwysing. MVAB het die beheer parameters aangepas om die spesifieke loonvrag in elke vlug te akkommodeer. Die vlugtoetse was suksesvol en die algoritme het die ossillasies, wat deur die loonvrag veroorsaak word, gedemp terwyl stabiele vlug gehandhaaf was.

Please refer to this item in SUNScholar by using the following persistent URL: http://hdl.handle.net/10019.1/107832
This item appears in the following collections: