The purpose of this thesis is to design a Model Predictive Control based Motion Planner unit for an Autonomous Vehicle. The unit should provide the trajectories of the inputs of the vehicle such that certain references are followed and, at the same time, fixed and moving obstacles are avoided safely. The first controller studied is an Active Front Steering system, where the input is only the steering angle. The controller is a linear model predictive control based on linearization of the nonlinear vehicle model and it was used to thoroughly investigate its limits, especially for what concerns the purpose of avoiding obstacles. This linear controller is then enhanced including also the control of the brake and throttle pedals. This allows us to further delineate properties and defects of the linear controller. Once the limits of the linear controller have been understood, the design of the nonlinear model predictive control, again dedicated to define the trajectories of the steering angle and of the brake and throttle pedal positions, has been carried out. We have then challenged it with fixed obstacles, with blind alleys and with moving obstacles. Also an extensive analysis on the controller sensitivity for what regards obstacle shapes, optimization algorithm and optimization initial conditions has been carried out in order to define the best setup.
Lo scopo di questa tesi è quello di progettare un’unità Motion Planner basata sul Model Predictive Control da poter essere utilizzata su un veicolo autonomo. L’unità dovrà provvedere a fornire le traiettoria dei controlli del veicolo in modo tale da permettere di seguire determinati riferimenti e, allo stesso tempo, di evitare ostacoli sia fissi che in movimento. Il primo controllore studiato è un sistema Active Front Steering, dove l’input del sistema è solamente l’angolo di sterzo. Il controllore è un Model Predictive Control lineare basato sulla linearizzazione del modello non lineare del veicolo, ed è stato usato per indagare a fondo i suoi limiti, specialmente per quanto riguarda la capacità di evitare ostacoli. Questo controllore lineare è stato quindi sviluppato in modo tale da includere anche il controllo del pedale del freno e dell’acceleratore. Questo ha permesso di definire più approfonditamente i pro e i contro di questo controllore. Infine si è progettato il controllore Model Predictive non lineare, volto a definire le traiettorie dell’angolo di sterzo e della posizione del pedale del freno e dell’acceleratore. Sono state quindi valutate le sue performance nei confronti di ostacoli fissi nello spazio, di vicoli ciechi e di ostacoli in movimento. Inoltre è stata condotta un’analisi dettagliata in funzione della forma dell’ostacolo, degli algoritmi di ottimizzazione e delle condizioni iniziali della ottimizzazione.
Model predictive control for an autonomous vehicle
DE VAL, NICOLA;FUSO, ANDREA
2012/2013
Abstract
The purpose of this thesis is to design a Model Predictive Control based Motion Planner unit for an Autonomous Vehicle. The unit should provide the trajectories of the inputs of the vehicle such that certain references are followed and, at the same time, fixed and moving obstacles are avoided safely. The first controller studied is an Active Front Steering system, where the input is only the steering angle. The controller is a linear model predictive control based on linearization of the nonlinear vehicle model and it was used to thoroughly investigate its limits, especially for what concerns the purpose of avoiding obstacles. This linear controller is then enhanced including also the control of the brake and throttle pedals. This allows us to further delineate properties and defects of the linear controller. Once the limits of the linear controller have been understood, the design of the nonlinear model predictive control, again dedicated to define the trajectories of the steering angle and of the brake and throttle pedal positions, has been carried out. We have then challenged it with fixed obstacles, with blind alleys and with moving obstacles. Also an extensive analysis on the controller sensitivity for what regards obstacle shapes, optimization algorithm and optimization initial conditions has been carried out in order to define the best setup.File | Dimensione | Formato | |
---|---|---|---|
2013_12_Deval_Fuso.pdf
Open Access dal 27/11/2014
Descrizione: Testo della tesi
Dimensione
2.36 MB
Formato
Adobe PDF
|
2.36 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/10589/87544