OPEN POSITIONS - MASTER THESIS

Application Deadline: 31 January 2020 

Interviews: first week of February

Apple grasping with robotic arm classic vs Reinforcement Learning

This thesis is aimed to compare traditional techniques and Deep Reinforcement Learning within the control (in grasping tasks) of a robotic arm.

Welcome Skills

ROS/Python/Tensorlfow

Supervisor:
M. Chiaberge, V. Mazzia, F. Salvetti, E. Sutera

Code: 2020PIC01

Vineyard extraction from UAV images

The thesis objective is to design and train a Deep Learning segmentation algorithm able to extract with a real time solution the binary map of a vineyard starting from drone imagery.

Welcome Skills

Python/Tensorflow/OpenCV

Supervisor:
M. Chiaberge, V. Mazzia, F. Salvetti

Code: 2020PIC02

Reinforcement Learning with camera and depth algorithm + Virtual framework

The thesis objective is to obtain an autonomous navigation system that makes only use of a monocamera and deep reinforcement algorithms. Firstly, the system will be trained on a virtual environment and later will be deployed on a real robot that should be able to autonomously navigate in an indoor environment.

Welcome Skills

Python/Tensorflow/OpenCV/RL

Supervisor:
M. Chiaberge, V. Mazzia, F. Salvetti, E. Sutera

Code: 2020PIC03

GPS/local navigation for precision agriculture

The thesis aims at integrating different technologies to get a fully autonomous UGV navigation system in agriculture contexts.

Welcome Skills

ROS/Python

Supervisor:
M. Chiaberge, V. Mazzia, F. Salvetti, D. Aghi, G. Fantin

Code: 2020PIC04

Precise indoor navigation, mapping and classification

The scope of this thesis is to adapt existing SLAM techniques to allow a Unmanned Ground Vehicle to navigate in an unknown indoor enviroment and progressively map it. The map then has to be analysed with a neural network to classify each room.

Welcome Skills

ROS/Python/SLAM

Supervisor:
M. Chiaberge, R. Cappellaro, A. Boschi, D. Aghi

Code: 2020PIC05

Person detection and tracking with IR/FIR cameras

The scope of this thesis is to identify and classify a person using images coming from InfraRed and Far InfraRed (thermal) cameras. These images will be processed using deep learning algorithms to detect and track a person in the frame.

Welcome Skills

Python/Tensorflow/Electronics

Supervisor:
M. Chiaberge, V. Mazzia, R. Cappellaro, A. Boschi

Code: 2020PIC06

Swarm flight with PX4-based quadcopters

Thesys on swarm flight, fleet of 2-4 quadcopters running PX4, need to modify QGroundControl for controlling multiple vehicles with the goal of a final demo.

Welcome Skills

C++/Geeks

Supervisor:
M. Chiaberge, S. Silvestro, G. Dara, G. Fantin

Code: 2020PIC07

PX4 customization, drivers, mixers

Thesys on PX4 open source flight stack environment, from data link to real time operating system, custon peripheral driving and control mixing for autopilot customization.

Welcome Skills

C++/Geeks/Control

Supervisor:
M. Chiaberge, S. Silvestro, G. Dara

Code: 2020PIC08

UAV/UGV control&navigation with smartphone

The thesis goal is to optimize an Android app implementing a drone autopilot with PID controller using smartphone sensors (IMU, GNSS, WiFi, Camera) and make it fly.

Welcome Skills

C++/java/Android

Supervisor:
M. Chiaberge, S. Silvestro, G. Dara, D. Aghi

Code: 2020PIC09

Motion planning integration for service robotic applications (no LIDAR if possible)

This thesis is aimed to the integration of systems for human assistance, such as human motion planning, pose estimation (for instance “walking”, “lying down”). Such information are then used for evalutaing whether there’s a dangerous situation and eventually give an alarm signal.

Welcome Skills

ROS/Python/OpenCV

Supervisor:
M. Chiaberge, A. Boschi, E. Sutera

Code: 2020PIC10

UWB: NLOS compensation with deep learning

UWB is one of the most promising technology for indoor localization. The typical deployment is made up of three (or more) fixed sensors called anchors. Measuring the distances between the anchors and the target device, it is possible to compute its position with centimeter level of accuracy. Nevertheless, the performances of the system quickly degrade when an obstacle blocks the line of sight between the anchors and the target, introducing an offset in the ranging measurements. The aim of this thesis is to create a proper database to train a deep-learning algorithm that is able to compensate in real time the offset caused by non-line of sight conditions.

Welcome Skills

Python/Tensorflow/Signal analysis

Supervisor:
M. Chiaberge, V. Mazzia, G. Fantin

Code: 2020PIC11

Anchor self-localization indoor/outdoor

UWB is one of the most promising technology for localization in GPS-denied environments. The typical deployment is made up of three (or more) fixed sensors called anchors. Measuring the distances between the anchors and the target device, it is possible to compute its position with centimeter level of accuracy. Whenever the system must be set up in a short time (e.g. emergency scenarios), the need to accurately measure the position of the anchors becomes a limitation. The aim of this thesis is to develop an effective algorithm to automatically compute the relative positions of the anchors knowing the relative distances between them. In outdoor scenarios, a possible extension is to equip the devices with GPS and fuse both measurements (UWB + GPS) to obtain the precise geolocation of the anchors. 

Welcome Skills

C++

Supervisor:
M. Chiaberge, G. Fantin

Code: 2020PIC12

Real-time communications between robots (drones/rovers) using mobile devices

The scope of this thesis is to study and implement a system to allow real-time communications between robots (drones/rovers) using mobile devices, such as smartphones. The focus is on extending an Android application to tightly control communication latency in order to transmit time-sensitive information such as positioning, as well as to eventually develop a simple communication protocol that allows direct communication between the devices even when access point support is missing.

Supervisor:
E. Masala

Collaboration:
~

Code: 2019COM01

Multimedia data processing from robots (drones/rovers)

The scope of this thesis is to study and implement a system that can efficiently handle multimedia data produced by drones and rovers, e.g. RGB or multispectral images. It is expected to develop efficient processing and storage algorithms, as well as to identify interesting areas in the image that might need further elaboration. Other types of information could include data coming from sensors of a typical smartphone device including, for instance, wireless signal strength.
A second goal is to highlight critical aspects of this kind of technology in robotic applications suggesting counterfeiting solutions.

Supervisor:
E. Masala

Collaboration:
~

Code: 2019COM02

IR images for visual odometry

The thesis will be focused in the testing and integration of an InfraRed sensor helpful for the autonomous navigation. This system will be composed by two IR active cameras and an RGB sensor, and it will be integrated in a visul odometry process to determine the position and orientation of a ground/aerial vehicle in a real environment.

Supervisor:
I. Aicardi, A. Lingua, V. Di Pietra

Collaboration:~

Code: 2019VIS01

Ultra wideband implementation on UAVs

The scope of the thesis is to integrate and test the Ultra Wideband technology on Unmanned Aerial Vehicles. Sensors and algorithms have already been tested in controlled and ground case studies. The thesis will be focused on sensors installation and management, analysis and mitigation of engine interferences, technology testing and validation.

Supervisor:
A. Lingua, V. Di Pietra, M. Chiaberge

Collaboration:
~

Code:  2019VIS02

Hyperspectral analysis in precision agriculture environment

Precision agriculture is now a key topic where the use of multi-bands sensors can help in the early detection of plant diseases and in the study of their state of health. The scope of this thesis is to evaluate the use of a hyperspectral camera in a vineyard environment and to study and implement its installation on board of an Unmanned Aerial Vechicle.

Supervisor: I. Aicardi, A. Lingua, M.A. Musci Collaboration:~ Code:  2019VIS03

Autopilot implementation on a ground vehicle

The scope of the thesis is the implementation of an autopilot on a ground vehicle. The vehicle is already available and the system must provide an autonomous (and programmed) navigation with the possibility to integrate sensors, like a First Person View system, positioning sensors, cameras and laser for points cloud generation.

Supervisor:
I. Aicardi, A. Lingua, P. Maschio

Collaboration:
~

Code: 2019ROB08