With the advent of low-cost electronic systems for the automation of certain func- tions related to the conduct of ight, many enthusiasts and scholars have had the opportunity to turn their model aircraft into real Unmanned Aerial Vehicle (UAV). These electronic systems are available on the market as "ready to y": only the integration phase and performing a calibration procedure are enough to obtain an aircraft capable of performing automatic missions controlling position, airspeed and altitude, eventually without the intervention of the operator on the ground. During our experiments, a xed-wing motor-glider was tted with a "Pixhawk" Flight Contro Unit (FCU), one of the many hardware available on the market for such applications. On this hardware, the ArduPilot rmware was installed, being already known in the environment of the Polytechnic University Of Milan. The Pixhawk was then equipped with the telemetry module needed to receive and send data to the ground in order to give to the operator real-time data of the UAV. These solutions are having a major impact in the current economy: many activi- ties can be carried out more e ectively and economically than traditional solutions and, not least, new professionals have also appeared. Compared to multirotors, xed-wing aircraft show the undeniable advantage of having greater range and endurance, as well as a greater payload capability. In this thesis, the main tasks necessary to tune the control laws inherent the FCU will be addressed to improve the performance of the attitude control, given the available hardware (servos, size and de ection of the aerodynamic surfaces, etc. . . ) because the automatic tun- ing that such FCU's can perform online during appropriate ight missions are of course e ective and reliable, but they certainly do not carry it out optimally nor even try to satisfy any user speci cation. The so-called "stock" tuning of the control laws obtained using the "Autotune" function available with ArduPilot, tunes only some gains of the attitude control loops during the dedicated tests, setting the remaining gains with pre-defned values based on the aggressiveness of the tuning desired by the user. The aim of the thesis is therefore to exclude the "Autotune" function from the UAV tuning procedure calibrating, through specifc experiments and computations, the above gains of the control loops to achieve the performance desired by the user. These performance constraints can be various: control, mission, atmospheric disturbance rejection performance, and others. In this thesis, this procedure was carried out by exploring the potentialof the Virtual Reference Feedback Tuning (VRFT) technique for calculating the gains of the control loops, combined with an iterative methodology to evaluate the obtained performance and to get a evaluation to the goodness of the closed- loop system with the user specifcations. An iterative procedure based on the exploration of a finite domain of different combinations of the reference model's parameters used in the VRFT procedure, made it possible to explore a set of pos- sible gain solutions. Those solutions have been inserted in the closed-loop system and the performance have been then evaluated through a cost function to find the best possible combination of parameters to control the system according to user's specifications. The computed performance was subsequently compared with the in- ight tests and the results, among with a detailed discussion, are presented in this thesis.
Con l'avvento dei sistemi elettronici a basso costo per lo svolgimento automatico di alcune funzioni legate alla conduzione del volo, numerosi appassionati e stu- diosi hanno avuto la possibilit a di trasformare il proprio aeromodello in vero e proprio Aeromobile a Pilotaggio Remoto (APR). Tali sistemi elettronici si presen- tano come "pronti al volo", ovvero e su ciente integrarli al proprio aeromodello ed eseguire alcune procedure di calibrazione per ottenere un velivolo in grado di eseguire missioni in maniera autonoma, gestendo posizione, velocit a e quota senza l'intervento dell'operatore a terra. Durante la nostra sperimentazione, un motoaliante ad ala ssa e stato arricchito di un modulo "Pixhawk", uno dei tanti hardware disponibili sul mercato per tali applicazioni. Su tale hardware e stato installato ArduPilot come rmware, essendo gi a noto nell'ambiente del Politec- nico di Milano. La Flight Control Unit (FCU) Pixhawk e stata quindi corredata del modulo telemetria necessario per la ricezione e l'invio dei dati a terra cos da fornire all'operatore i dati in tempo reale delle prove in atto. Tali soluzioni stanno avendo un grande impatto nell'economia attuale: numerose attivit a infatti possono essere svolte in maniera pi u e cace ed economica rispetto alle soluzioni tradizionali e, non da ultimo, ha visto comparire anche nuove gure professionali. Rispetto ai multirotori, i velivoli ad ala ssa mostrano l'innegabile vantaggio di poter disporre di una maggiore autonomia (sia oraria che chilometrica) nonch e disporre di un maggior carico utile. In questa tesi verranno a rontati i principali temi necessari alla calibrazione delle leggi di controllo insite nel computer di bordo per migliorare le prestazioni sul controllo di assetto dato l'hardware a disposizione (servocomandi, dimensioni e movimento delle super ci mobili, ecc...) in quanto le calibrazioni che tali unit a di controllo possono svolgere on-line durante appo- site missioni di volo svolgono in maniera accettabile tale compito ma sicuramente non lo portano a compimento in maniera ottimale. La calibrazione cosiddetta "stock", ovvero la calibrazione delle leggi di controllo ottenuta usando la funzione "Autotune" disponibile con ArduPilot, regola solamente alcuni guadagni degli anelli di controllo dell'assetto durante le prove dedicate, impostando i rimanenti con valori prede niti a seconda dell'aggressivit a della traratura scelta dall'utente. L'obiettivo della tesi e quindi escludere la funzione "Autotune" dalla messa in linea dell'APR e tarare, mediante speci ci esperimenti ed elaborazioni, i suddetti guadagni degli anelli di controllo per ottenere le prestazioni desiderate dall'utente.Tali vincoli sulle prestazioni possono essere di varia natura: prestazioni di con- trollo, di missione, di rigetto dei disturbi atmosferici e altri. In questa tesi si e esplorata tale procedura mediante l'uso della tecnica VRFT (Virtual Reference Feedback Tuning) per il calcolo dei guadagni degli anelli di controllo, successiva- mente inseriti nel sistema in anello chiuso e valutate le prestazioni. Una proce- dura iterativa basata sulla variazione dei parametri dei modelli di riferimento in uso nella procedura VRFT entro range pre ssati ha reso possibile esplorare un insieme di possibili soluzioni, alla ricerca della soluzione ottima per il soddisfaci- mento delle speci che minimizzando determinati indici di merito. Le prestazioni calcolate per via numerica sono state successivamente comparate con le prove in volo ed i risultati delle stesse presentati in questa tesi.
Modelling, identification and control of a fixed-wing UAV
GATTI, ANDREA
2018/2019
Abstract
With the advent of low-cost electronic systems for the automation of certain func- tions related to the conduct of ight, many enthusiasts and scholars have had the opportunity to turn their model aircraft into real Unmanned Aerial Vehicle (UAV). These electronic systems are available on the market as "ready to y": only the integration phase and performing a calibration procedure are enough to obtain an aircraft capable of performing automatic missions controlling position, airspeed and altitude, eventually without the intervention of the operator on the ground. During our experiments, a xed-wing motor-glider was tted with a "Pixhawk" Flight Contro Unit (FCU), one of the many hardware available on the market for such applications. On this hardware, the ArduPilot rmware was installed, being already known in the environment of the Polytechnic University Of Milan. The Pixhawk was then equipped with the telemetry module needed to receive and send data to the ground in order to give to the operator real-time data of the UAV. These solutions are having a major impact in the current economy: many activi- ties can be carried out more e ectively and economically than traditional solutions and, not least, new professionals have also appeared. Compared to multirotors, xed-wing aircraft show the undeniable advantage of having greater range and endurance, as well as a greater payload capability. In this thesis, the main tasks necessary to tune the control laws inherent the FCU will be addressed to improve the performance of the attitude control, given the available hardware (servos, size and de ection of the aerodynamic surfaces, etc. . . ) because the automatic tun- ing that such FCU's can perform online during appropriate ight missions are of course e ective and reliable, but they certainly do not carry it out optimally nor even try to satisfy any user speci cation. The so-called "stock" tuning of the control laws obtained using the "Autotune" function available with ArduPilot, tunes only some gains of the attitude control loops during the dedicated tests, setting the remaining gains with pre-defned values based on the aggressiveness of the tuning desired by the user. The aim of the thesis is therefore to exclude the "Autotune" function from the UAV tuning procedure calibrating, through specifc experiments and computations, the above gains of the control loops to achieve the performance desired by the user. These performance constraints can be various: control, mission, atmospheric disturbance rejection performance, and others. In this thesis, this procedure was carried out by exploring the potentialof the Virtual Reference Feedback Tuning (VRFT) technique for calculating the gains of the control loops, combined with an iterative methodology to evaluate the obtained performance and to get a evaluation to the goodness of the closed- loop system with the user specifcations. An iterative procedure based on the exploration of a finite domain of different combinations of the reference model's parameters used in the VRFT procedure, made it possible to explore a set of pos- sible gain solutions. Those solutions have been inserted in the closed-loop system and the performance have been then evaluated through a cost function to find the best possible combination of parameters to control the system according to user's specifications. The computed performance was subsequently compared with the in- ight tests and the results, among with a detailed discussion, are presented in this thesis.| File | Dimensione | Formato | |
|---|---|---|---|
|
thesis_main.pdf
accessibile in internet per tutti
Dimensione
3.04 MB
Formato
Adobe PDF
|
3.04 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/148374