Legged robots, mainly quadrupeds, saw a significant increase of interest by the scientific community in the last years due to their versatility, adaptability and potential for real- world application, even if they present several control challenges due to their intrinsic underactuation. The today state of the art robots, often equipped with sensors and manipulators, are able to perform complicated tasks also on rough and unpredictable terrains, notwithstanding this, there is wide room for improvement both from the control point of view and also regarding path planning and sensor fusion. The reason behind this thesis is to design and simulate the full low level control scheme of a light and small quadruped robot, side by side with learning a control framework for its implementation. Namely, it is not taken into account the high level controller, the reason behind its movement, but the controller will address the problem of taking as input a velocity along the x-y plane and giving as an output the torque control for each of the joints. Along with this, the controller will take into account the robot stability, the non slipping condition of the legs and wrench optimization. The proposed model and the low level whole-body controller have been successfully used to to simulate tasks such as walking in different surface conditions and performing pushups. From the implementation point of view, the controller has been coded and the robot simu- lated and, analyzing the libraries used in the control framework, it has been proposed the use of a more effective and easy to use C++ library to compute the Rigid Body Dynamics algorithms.
I robot quadrupedi hanno visto un significativo aumento di interesse da parte della co- munità scientifica negli ultimi anni grazie alla loro versatilità, adattabilità e potenzialità di applicazione nel mondo reale, anche se presentano diverse sfide di controllo a causa della loro intrinseca sottoattuazione. I robot allo stato dell’arte, spesso dotati di sensori e manipolatori, sono in grado di svolgere compiti complicati anche su terreni accidentati e imprevedibili, tuttavia esistono ampi margini di miglioramento sia dal punto di vista del controllo sia per quanto riguarda la pianificazione del percorso e l’integrazione dei sensori. Lo scopo di questa tesi è progettare e simulare lo schema di controllo a basso livello di un robot quadrupede leggero e di piccole dimensioni, parallelamente all’apprendimento di un framework di controllo per la sua implementazione. In altre parole, non viene preso in considerazione il controllore di alto livello, la ragione del suo movimento, ma il controllore di basso livello affronterà il problema di prendere in ingresso una velocità lungo il piano x-y e dare in uscita il controllo della coppia per ciascuno dei giunti. Inoltre, il controllore terrà conto della stabilità del robot, della condizione di non slittamento delle zampe e dell’ottimizzazione delle forze che agiscono sul tronco. Il modello proposto e il controllore a basso livello sono stati utilizzati per simulare compiti come camminare in diverse condizioni di superficie ed eseguire flessioni. Dal punto di vista dell’implementazione, sono stati realizzati sia il modello che il controllore e, analizzando le librerie utilizzate dal framework, è stato proposto l’uso di una libreria C++ più efficace e facile da usare per calcolare gli algoritmi di dinamica del corpo rigido.
Low level whole-body controller for a quadruped robot - design and implementation
Di Francesco, Daniele
2022/2023
Abstract
Legged robots, mainly quadrupeds, saw a significant increase of interest by the scientific community in the last years due to their versatility, adaptability and potential for real- world application, even if they present several control challenges due to their intrinsic underactuation. The today state of the art robots, often equipped with sensors and manipulators, are able to perform complicated tasks also on rough and unpredictable terrains, notwithstanding this, there is wide room for improvement both from the control point of view and also regarding path planning and sensor fusion. The reason behind this thesis is to design and simulate the full low level control scheme of a light and small quadruped robot, side by side with learning a control framework for its implementation. Namely, it is not taken into account the high level controller, the reason behind its movement, but the controller will address the problem of taking as input a velocity along the x-y plane and giving as an output the torque control for each of the joints. Along with this, the controller will take into account the robot stability, the non slipping condition of the legs and wrench optimization. The proposed model and the low level whole-body controller have been successfully used to to simulate tasks such as walking in different surface conditions and performing pushups. From the implementation point of view, the controller has been coded and the robot simu- lated and, analyzing the libraries used in the control framework, it has been proposed the use of a more effective and easy to use C++ library to compute the Rigid Body Dynamics algorithms.File | Dimensione | Formato | |
---|---|---|---|
tesi_v0.5.3_DiFrancesco.pdf
Open Access dal 14/03/2025
Dimensione
3.3 MB
Formato
Adobe PDF
|
3.3 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/10589/219721