This thesis presents a collision-free switching model predictive control (SMPC) framework for the aggregation of quadrotor unmanned aerial vehicles (UAVs). The study begins with modeling the kinematics and dynamics of quadrotors, followed by the design of low-level controllers. Such controllers constitute the inner layer of the hierarchical architecture adopted in this thesis: they are responsible of controlling the actuators of the drone, in this case the rotor motors, to follow the high-level commands, which are the position and the attitude provided by the outer loop. Two different approaches are used to design the controllers. The first one involves linearizing the system around the hovering condition, that is the equilibrium condition in which the quadrotor has constant height and zero longitudinal speed. The second approach is to obtain the equivalent linear system from the nonlinear model using feedback linearization, by the application of differential flatness theory, and then the regulator is designed using a pole placement approach. The high-level control is achieved using a SMPC methodology, which reduces the complexity associated with the quadrotor’s high degrees of freedom by selecting motions from a finite set. The proposed SMPC includes constraints to take into account collision-free trajectories both among quadrotors, and with obstacles in the environment. To this purpose, different techniques to obtain linear constraints are proposed. Specifically, the obstacles are first approximated as spherical objects, then the spherical obstacle assumption is relaxed, and two different techniques for approximating arbitrarily shaped obstacles are introduced: covering the obstacle’s volume with spheres, namely sphere packing, and approximating the obstacle by computing its convex hull. The methods discussed in this thesis are then applied to different simulated 3D environments. The scalability and feasibility of the strategies proposed in this thesis are finally also discussed.
L’obiettivo della tesi è la realizzazione di un algoritmo switching model predictive control (SMPC) per l’aggregazione di quadricotteri senza pilota (UAV), che tenga conto di vincoli per l’evitamento di collisioni. Il primo contributo della tesi è la modellazione della cinematica e della dinamica dei quadrirotori, seguita dalla progettazione dei controllori di basso livello. I controllori di basso livello costituiscono il livello interno dell’architettura gerarchica adottata. Essi sono responsabili del controllo degli attuatori del drone, in questo caso i motori dei rotori, per inseguire i riferimenti generati dal livello superiore dello schema, ovvero la posizione e l’assetto forniti dall’anello esterno. Due diversi approcci sono poi utilizzati per progettare i controllori. Il primo prevede la linearizzazione del sistema attorno alla condizione di hovering, ovvero la condizione di equilibrio in cui il quadricottero ha altezza costante e velocità longitudinale nulla. Il secondo approccio consiste nell’ottenere il sistema lineare equivalente dal modello non lineare utilizzando un metodo di feedback linearization, ottenuta applicando la teoria della differential flatness, e progettando successivamente il regolatore con un approccio di posizionamento dei poli. Il controllo di alto livello è costituito da un algoritmo SMPC, che riduce la complessità associata agli elevati gradi di libertà del modello del quadricottero, selezionando un movimento tra quelli presenti in un insieme predefinito. L’algoritmo SMPC proposto include vincoli per garantire il movimento dei quadricotteri senza collisioni, sia con altri quadricotteri, sia con ostacoli presenti nell’ambiente. A tal fine, vengono proposte diverse tecniche per esprimere i vincoli tramite disequazioni lineari. Nello specifico, gli ostacoli sono inizialmente approssimati come oggetti sferici. L’assunzione degli ostacoli sferici viene successivamente rilassata, e vengono introdotte due diverse tecniche per approssimare ostacoli di forma qualsiasi: coprire il volume dell’ostacolo con sfere, noto come sphere packing, e approssimare l’ostacolo calcolandone l’inviluppo convesso. I metodi discussi in questa tesi sono stati quindi verificati in diversi ambienti 3D simulati. La tesi si conclude con una discussione relativa alla scalabilità e alla realizzabilità pratica delle strategie proposte.
Collision-free switching model predictive control for quadrotor UAVs aggregation
Nicolò, Matteo
2023/2024
Abstract
This thesis presents a collision-free switching model predictive control (SMPC) framework for the aggregation of quadrotor unmanned aerial vehicles (UAVs). The study begins with modeling the kinematics and dynamics of quadrotors, followed by the design of low-level controllers. Such controllers constitute the inner layer of the hierarchical architecture adopted in this thesis: they are responsible of controlling the actuators of the drone, in this case the rotor motors, to follow the high-level commands, which are the position and the attitude provided by the outer loop. Two different approaches are used to design the controllers. The first one involves linearizing the system around the hovering condition, that is the equilibrium condition in which the quadrotor has constant height and zero longitudinal speed. The second approach is to obtain the equivalent linear system from the nonlinear model using feedback linearization, by the application of differential flatness theory, and then the regulator is designed using a pole placement approach. The high-level control is achieved using a SMPC methodology, which reduces the complexity associated with the quadrotor’s high degrees of freedom by selecting motions from a finite set. The proposed SMPC includes constraints to take into account collision-free trajectories both among quadrotors, and with obstacles in the environment. To this purpose, different techniques to obtain linear constraints are proposed. Specifically, the obstacles are first approximated as spherical objects, then the spherical obstacle assumption is relaxed, and two different techniques for approximating arbitrarily shaped obstacles are introduced: covering the obstacle’s volume with spheres, namely sphere packing, and approximating the obstacle by computing its convex hull. The methods discussed in this thesis are then applied to different simulated 3D environments. The scalability and feasibility of the strategies proposed in this thesis are finally also discussed.File | Dimensione | Formato | |
---|---|---|---|
[Executive_summary Nicolò][23_24]Collision_avoidance_SMPC_UAVs.pdf
solo utenti autorizzati a partire dal 01/07/2027
Descrizione: Executive Summary
Dimensione
704.02 kB
Formato
Adobe PDF
|
704.02 kB | Adobe PDF | Visualizza/Apri |
[Tesi Nicolò][23-24]Collision_avoidance_SMPC_UAVs.pdf
solo utenti autorizzati a partire dal 01/07/2027
Descrizione: Thesis
Dimensione
18.82 MB
Formato
Adobe PDF
|
18.82 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/10589/223777