The purpose of this thesis is to develop strategies for autonomous in-space assembly of a large scale exible robotic multibody system, e ciently launched and deployed to compose a dynamical recon gurable structure. The whole problem is investigated: formation control and guidance of the deployed agents, proximity control of the single exible spacecraft for precise positioning and vibrations suppression, and stabilization of the obtained assembled structure. The control problem at the marco-scale involves formation ying of a multiple agent distributed system, in which deterministic control techniques are applied to guide and recon gure the agents according to the speci ed mission. The swarm of exible spacecrafts is rst modeled as a set of rigid bodies in orbital space environment. Potential Fields approaches allow to maintain the formation's con guration following the leading agent, alongside with a collision avoidance strategy. These methods are obtained de ning virtual force elds that a ect each agent and let them interact with the surroundings. On a smaller scale, the proximity control of the spacecraft is managed with an output consensus controller combined with a re ned collision avoidance strategy, assuming the acquisition of the adjacent agents' output. The problem is also addressed with a sub-optimal solution approach, the Inverse Dynamics Virtual Domain (IDVD) method, which sacri ces optimality assessing improvements in terms of precision and robustness with respect to external disturbances. The output states of these control strategies are feeded as inputs to a Negative Derivative Feedback (NDF) controller to suppress the vibrations of the exible spacecraft's appendage. To prevent vibrations from a ecting the guidance and control of the single agent at the macro-scale, the NDF controller and the proximity control technique must be run in parallel. The proposed NDF controller successfully damps the vibrations of the exible appendages of the system so that they do not interfere with its maneuvers and standard approaches for ridid docking can be used. Then, concerning the assembly aspect, a fairly simple analytical docking model is presented to describe the transition from the formation of agents to the assembled structure. The ultimate goal is to check if the control logic is capable of managing the maneuver and indeed, it does. Each time a docking between the main structure and an agent is performed, a dynamical re-allocation of sensors and actuators is performed in order to have a better performance of the control logic. Exploiting the studies done in adaptive optics, a way to decouple the actuators of the assembly is presented. In the end some numerical simulations are done, with three possible designs for the single agent. The study shows how this choice highly a ects the behaviour of the assembled structure. The better choice is related to the ultimate application and situation of the problem that wants to be addressed.
Lo scopo di questa tesi e quello di sviluppare un sistema di assemblaggio autonomo nello spazio di uno sciame di agenti essibili, capaci di con gurarsi in un'unica struttura ricon gurabile. Lo studio si divide in tre parti principali: guida e controllo dei singoli agenti; controllo di prossimit a e studio della essibilit a dell'agente per la soppressione delle vibrazioni e posizionamento di precisione; stabilizzazione e logica di controllo della struttura assemblata. Il problema a livello macroscopico e la formazione della otta composta da un sistema distribuito di molteplici agenti. Sono applicate tecniche di controllo deterministiche per guidare e ricon gurare la formazione di volo degli agenti in base alla speci ca missione. Inizialmente lo sciame e modellato con una serie di corpi rigidi che orbitano nello spazio intorno allo terra. In particolare e usata la tecnica dei campi potenziali arti ciali che permette il mantenimento della formazione durante le manovre, seguendo un agente leader. In parallelo viene lanciato un algoritmo per evitare la collisione degli agenti. A livello microscopico, il controllo di prossimit a tra i satelliti e sviluppato tramite una strategia di "output consensus" in parallelo ad un pi u ra nato algoritmo per evitare la collisione tra agenti che prende in considerazione la geometria degli stessi e la loro intrinseca essibilit a. Lo stesso problema e stato approcciato anche con il metodo IDVD (Inverse Dynamics Virtual Domain), che con una soluzione subottimale garantisce una soluzione precisa e robusta ai disturbi. Inoltre risulta essere una tecnica estremamente e ciente e leggera dal punto di vista computazionale. Le manovre di ogni agente sono date in input ad un modello dell'appendice essibile del satellite stesso che ne genera l'andamento delle vibrazioni. Una logica di controllo in risonanza, il Negative Derivative Feedback, e usata per lo smorzamento. Grazie allo smorzamento consistente ed e cace, le manovre possono essere è effettuate come in una normale situazione di agente rigido. E' stato successivamente sviluppato un modello analitico della fase relativa all'aggancio degli agenti. Essendo analitico non ha la pretesa di essere realistico, e servito a vedere come rispondano le logiche di controllo per lo smorzamento delle vibrazioni all'improvviso cambio di parametri del sistema. Ogni volta che un agente si assembla alla struttura, avviene una ricollocazione automatica dei sensori e attuatori per permettere una maggiore e cienza dei sistemi di controllo. Sfruttando gli studi sulle ottiche adattive, e stato proposto un modo per ottenere il disaccoppiamento tra gli attuatori e sensori della struttura. In ne diverse simulazioni numeriche sono state eseguite con tre diverse proposte di design per il singolo agente. E' stato visto come il comportamento della struttura assemblata è altamente dipendente dalla geometria del singolo agente, da cui dipende il modo in cui avviene l'aggancio.
Autonomous in-space assembly : modeling and control
PAPARELLA, FABIO;
2018/2019
Abstract
The purpose of this thesis is to develop strategies for autonomous in-space assembly of a large scale exible robotic multibody system, e ciently launched and deployed to compose a dynamical recon gurable structure. The whole problem is investigated: formation control and guidance of the deployed agents, proximity control of the single exible spacecraft for precise positioning and vibrations suppression, and stabilization of the obtained assembled structure. The control problem at the marco-scale involves formation ying of a multiple agent distributed system, in which deterministic control techniques are applied to guide and recon gure the agents according to the speci ed mission. The swarm of exible spacecrafts is rst modeled as a set of rigid bodies in orbital space environment. Potential Fields approaches allow to maintain the formation's con guration following the leading agent, alongside with a collision avoidance strategy. These methods are obtained de ning virtual force elds that a ect each agent and let them interact with the surroundings. On a smaller scale, the proximity control of the spacecraft is managed with an output consensus controller combined with a re ned collision avoidance strategy, assuming the acquisition of the adjacent agents' output. The problem is also addressed with a sub-optimal solution approach, the Inverse Dynamics Virtual Domain (IDVD) method, which sacri ces optimality assessing improvements in terms of precision and robustness with respect to external disturbances. The output states of these control strategies are feeded as inputs to a Negative Derivative Feedback (NDF) controller to suppress the vibrations of the exible spacecraft's appendage. To prevent vibrations from a ecting the guidance and control of the single agent at the macro-scale, the NDF controller and the proximity control technique must be run in parallel. The proposed NDF controller successfully damps the vibrations of the exible appendages of the system so that they do not interfere with its maneuvers and standard approaches for ridid docking can be used. Then, concerning the assembly aspect, a fairly simple analytical docking model is presented to describe the transition from the formation of agents to the assembled structure. The ultimate goal is to check if the control logic is capable of managing the maneuver and indeed, it does. Each time a docking between the main structure and an agent is performed, a dynamical re-allocation of sensors and actuators is performed in order to have a better performance of the control logic. Exploiting the studies done in adaptive optics, a way to decouple the actuators of the assembly is presented. In the end some numerical simulations are done, with three possible designs for the single agent. The study shows how this choice highly a ects the behaviour of the assembled structure. The better choice is related to the ultimate application and situation of the problem that wants to be addressed.File | Dimensione | Formato | |
---|---|---|---|
Tesi_Paparella_Zanutto.pdf
non accessibile
Descrizione: Testo della Tesi
Dimensione
6.96 MB
Formato
Adobe PDF
|
6.96 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/10589/152997