The work presented in this thesis is in the area of autonomous robotic exploration of an unknown environment. The general purpose is to develop robots able to move autonomously and effectively in environments in order to build their representations, or maps. The main context of application of this work is the USAR (Urban Search And Rescue), where a robot has the task of searching for victims and survivors in an environment affected by earthquakes, collapses, fires, or other disasters. In these cases, the usage of autonomous or semi-autonomous robots allows to limit the risks deriving from a direct intervention of the human rescuers, preserving their lives. The purpose of this thesis is to investigate a new concept of solution for the exploration: while usually it is preferred to send the robot to locations that maximize the information gain over the unknown portions of the environment, in this work we analyse the effect of sending the robot to locations that minimize it. The intuition is that, when the objective is to completely explore an unknown environment, maximising the information gain often implies leaving behind corners and other small portions of the environment that will have to be reached later, traversing already known areas; minimizing the information gain, instead, these little portions of environment would be explored first, avoiding to have to return back in order to complete the exploration and so saving some time. The experimental results of this thesis seem to show that this approach can bring some benefits in the majority of the cases we considered and that it could increase the performance of the robots to decrease the exploration time.
Il lavoro svolto in questa tesi si colloca nell’ambito dell’esplorazione robotica autonoma di un ambiente sconosciuto. Lo scopo generale è quello di costruire robot che siano in grado di muoversi autonomamente ed efficacemente in ambienti inizialmente sconosciuti al fine di costruire una loro rappresentazione, o mappa. Il contesto principale di applicazione di questo lavoro è l’USAR (Urban Search And Rescue), nel quale un robot ha il compito di ricercare vittime e superstiti all’interno di un ambiente colpito da terremoti, crolli, incendi o altri disastri. In tali casi l’utilizzo di robot autonomi o semi-autonomi permette di limitare i rischi derivanti da un intervento diretto dei soccorritori, preservando le loro vite. Scopo di questa tesi è indagare un nuovo concetto di soluzione per l'esplorazione: mentre solitamente si preferisce spostare il robot in posizioni che massimizzano l’informazione ottenuta sull’ambiente ancora sconosciuto, in questo lavoro si analizza un approccio che preferisce per il robot posizioni che la minimizzano. L’intuizione è che, quando si vuole esplorare un ambiente in ogni sua parte, massimizzare l’informazione significa spesso lasciare indietro angoli o piccole porzioni di ambiente che andranno visitati in seguito, attraversando nuovamente aree già conosciute; minimizzando l’informazione ottenuta, invece, si esplorerebbero prima queste piccole porzioni di ambiente, evitando di dovere tornare indietro per completare l’esplorazione e risparmiando così del tempo. I risultati sperimentali ottenuti nella tesi sembrano mostrare che questo approccio possa apportare benefici nella maggioranza dei casi considerati e che quindi sia un possibile modo per migliorare le prestazioni del robot e diminuire i tempi di esplorazione.
Looking for less informative frontiers in autonomous robotic exploration : an experimental account
BAGNA, FRANCESCO MATTEO
2017/2018
Abstract
The work presented in this thesis is in the area of autonomous robotic exploration of an unknown environment. The general purpose is to develop robots able to move autonomously and effectively in environments in order to build their representations, or maps. The main context of application of this work is the USAR (Urban Search And Rescue), where a robot has the task of searching for victims and survivors in an environment affected by earthquakes, collapses, fires, or other disasters. In these cases, the usage of autonomous or semi-autonomous robots allows to limit the risks deriving from a direct intervention of the human rescuers, preserving their lives. The purpose of this thesis is to investigate a new concept of solution for the exploration: while usually it is preferred to send the robot to locations that maximize the information gain over the unknown portions of the environment, in this work we analyse the effect of sending the robot to locations that minimize it. The intuition is that, when the objective is to completely explore an unknown environment, maximising the information gain often implies leaving behind corners and other small portions of the environment that will have to be reached later, traversing already known areas; minimizing the information gain, instead, these little portions of environment would be explored first, avoiding to have to return back in order to complete the exploration and so saving some time. The experimental results of this thesis seem to show that this approach can bring some benefits in the majority of the cases we considered and that it could increase the performance of the robots to decrease the exploration time.| File | Dimensione | Formato | |
|---|---|---|---|
|
2018_12_Bagna.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: Thesis text
Dimensione
4.28 MB
Formato
Adobe PDF
|
4.28 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/10589/144744