Date of Award


Degree Type


Degree Name

Doctor of Philosophy in Electrical Engineering


Electrical, Computer, and Biomedical Engineering

First Advisor

Paolo Stegagno


In this work, we address and propose solutions to the problem of localization in single and multi-robot systems using vision as the sole sensor. We first propose a study on landmark identification as a step towards a localization setup for real-world robotic swarms setup. In real world, landmark identification is often tackled as a place recognition problem through the use of computationally intensive Convolution Neural Networks. However, the components of a robotic swarm usually have limited computational and sensing capabilities that allows only for the application of relatively shallow networks that results in large percentage of recognition errors. In [1], a cooperative approach was proposed where the authors assumed that all the robots were looking at the same landmark. Therefore, we initially proposed the use of a weighting factor to relapse this assumption. Through the use of simulation data, we showed that our approach provided high recognition rates even in situations in which the robots would look at different objects.

Secondly, we propose a landmark-based map localization system for robotic swarms. The system leverages the capabilities of the distributed landmark identification algorithm developed for robotic swarms presented above. The output of the landmark identification consists of a vector of probabilities that each individual robot is looking at a particular landmark in the environment. This vector was then used individually by each component of the swarm to feed the measurement update of a particle filter to estimate the robot location. The localization system was tested in simulation to validate its performance.

Thirdly, we propose a Monte Carlo approach to place recognition and landmark identification based indoor robot localization. Place recognition based localization systems in literature usually rely on a database of hundreds or even thousands of known locations. We have designed a particle filter that is capable of estimating a robots position based on the classification of the room in which the robot is. This means that, in case of a regular house environment, rather than using hundreds or thousands of known locations, the place recognition system will only need to be able to categorize the collected images into five-ten categories. Consequently, the map representation is very lightweight and does not have particular accuracy requirements, as many details about the environment can be neglected. Furthermore, room classification measurements are complemented with non-unique landmark identification measurements. As we recognize that identification and place recognition algorithms are not perfect, we also present a simulation study to assess the performance of the system with different correct recognition rates. Finally, we take the given system and examine the systems performance in a real experiment and present the results.

Lastly, we then propose a observability analysis on a simplified problem where only place recognition measurements are considered applied to a single cell. We show that the state is at least partially observable on the boundary of the cell, and that the entire state is only observable in the four corners.



To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.