This thesis aims to present a novel framework for the distributed source-seeking problem that takes advantage of the well-known divide-and-conquer paradigm. In particular, when deploying a team of robots in an unknown environment, the goal is to find as many sources as possible, a setting commonly known as multi-agent multi-source seeking. In the algorithms proposed in this study, each agent will take care only of a pre-assigned section of the area of interest, trying to avoid other types of collaboration and thus making the framework clearly distributed. Two main approaches are considered: offline and online, along with some of their variations. In the offline case, agents estimate the sources distribution collecting measurements along a predefined path, determined before the execution of the task. On the other hand, the online approach requires each agent to determine dynamically their next steps to accomplish the task. Those approaches are tested in a simulated environment and are compared against each other and state-of-the-art source seeking algorithms.
Questa tesi si propone di presentare un nuovo framework per risolvere il problema della ricerca distribuita di sorgenti, sfruttando il noto paradigma divide-et-impera. In particolare, utilizzando un team di robot in un ambiente completamente sconosciuto, l’obiettivo è individuare il maggior numero possibile di sorgenti. Il problema è comunemente noto in letteratura come multi-agent multi-source seeking. Negli algoritmi proposti in questo studio, ogni agente si occuperà unicamente di una sezione preassegnata dell’area di interesse, cercando di evitare altre forme di collaborazione e rendendo così il framework chiaramente distribuito. Sono stati considerati due approcci principali: offline e online, insieme ad alcune loro varianti. Nel caso offline, ciascun agente stima la distribuzione delle sorgenti con misurazioni effettuate lungo traiettorie predefinite, calcolate prima dell'esecuzione del task. D'altro canto, nell'approccio online ogni agente determina dinamicamente i propri passi per svolgere il task. Gli approcci proposti sono testati in un ambiente simulato e confrontati sia tra di loro che con altri algoritmi di source-seeking presenti in letteratura.
Multi-agent strategies for distributed multi-source seeking
CALIFANO, STEFANO
2024/2025
Abstract
This thesis aims to present a novel framework for the distributed source-seeking problem that takes advantage of the well-known divide-and-conquer paradigm. In particular, when deploying a team of robots in an unknown environment, the goal is to find as many sources as possible, a setting commonly known as multi-agent multi-source seeking. In the algorithms proposed in this study, each agent will take care only of a pre-assigned section of the area of interest, trying to avoid other types of collaboration and thus making the framework clearly distributed. Two main approaches are considered: offline and online, along with some of their variations. In the offline case, agents estimate the sources distribution collecting measurements along a predefined path, determined before the execution of the task. On the other hand, the online approach requires each agent to determine dynamically their next steps to accomplish the task. Those approaches are tested in a simulated environment and are compared against each other and state-of-the-art source seeking algorithms.| File | Dimensione | Formato | |
|---|---|---|---|
|
Califano_Stefano.pdf
accesso aperto
Dimensione
10.6 MB
Formato
Adobe PDF
|
10.6 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/96059