Available MSc Thesis topics
Modeling, control and optimization of Airborne Wind Energy Systems
We have several topics available for MSc theses on modeling, control and optimization of airborne wind energy systems. Please contact Lorenzo Fagiano for details.
Completed MSc Theses
1.
Scomazzon, Enrico
Closed Loop Identification and Simulation of an Octocopter Drone Masters Thesis
Automation and Control Engineering, Politecnico di Milano, 2021, (Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Danilo Saccani).
@mastersthesis{nokeyd,
title = {Closed Loop Identification and Simulation of an Octocopter Drone},
author = {Enrico Scomazzon},
year = {2021},
date = {2021-12-21},
address = {Politecnico di Milano},
school = {Automation and Control Engineering},
abstract = {In this thesis, we discuss the closed-loop identification and simulation of a commercial octocopter drone. Firstly, the forces acting on the drone are studied to derive a model of the vehicle.
Then, the control structure of an open-source commercial flight controller is integrated with the drone model. Such a control structure is different from the one employed on the real drone under study, a DJI S1000 model with A3 flight controller, which is not accessible. The main focus of the thesis is a novel closed-loop identification routine based on Simulation Error Method (SEM) that uses as optimization variables the drone physical parameters and the bandwidth of the cascaded loops describing the low-
level controller. This approach allows one to reduce the problem complexity by decreasing the number of optimization variables and at the same time helps to meet requirements regarding the robustness of the loops. Then, the performance of the identification procedure and the simulation accuracy is compared with data from experimental tests. This
thesis is concluded by including the identified model of both vehicle and controller in ROS-Gazebo simulation environment.},
note = {Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Danilo Saccani},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
In this thesis, we discuss the closed-loop identification and simulation of a commercial octocopter drone. Firstly, the forces acting on the drone are studied to derive a model of the vehicle.
Then, the control structure of an open-source commercial flight controller is integrated with the drone model. Such a control structure is different from the one employed on the real drone under study, a DJI S1000 model with A3 flight controller, which is not accessible. The main focus of the thesis is a novel closed-loop identification routine based on Simulation Error Method (SEM) that uses as optimization variables the drone physical parameters and the bandwidth of the cascaded loops describing the low-
level controller. This approach allows one to reduce the problem complexity by decreasing the number of optimization variables and at the same time helps to meet requirements regarding the robustness of the loops. Then, the performance of the identification procedure and the simulation accuracy is compared with data from experimental tests. This
thesis is concluded by including the identified model of both vehicle and controller in ROS-Gazebo simulation environment.
Then, the control structure of an open-source commercial flight controller is integrated with the drone model. Such a control structure is different from the one employed on the real drone under study, a DJI S1000 model with A3 flight controller, which is not accessible. The main focus of the thesis is a novel closed-loop identification routine based on Simulation Error Method (SEM) that uses as optimization variables the drone physical parameters and the bandwidth of the cascaded loops describing the low-
level controller. This approach allows one to reduce the problem complexity by decreasing the number of optimization variables and at the same time helps to meet requirements regarding the robustness of the loops. Then, the performance of the identification procedure and the simulation accuracy is compared with data from experimental tests. This
thesis is concluded by including the identified model of both vehicle and controller in ROS-Gazebo simulation environment.
2.
Manzo, Laura
Machine learning-based monitoring and diagnosis of industrial devices through current measurements Masters Thesis
Automation and Control Engineering, Politecnico di Milano, 2021, (Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Dr. Marco Lauricella).
@mastersthesis{nokey,
title = { Machine learning-based monitoring and diagnosis of industrial devices through current measurements},
author = {Laura Manzo},
url = {http://hdl.handle.net/10589/174149},
year = {2021},
date = {2021-04-28},
urldate = {2021-04-28},
journal = {Politecnico di Milano},
volume = {1},
number = {100},
pages = {43},
address = {Politecnico di Milano},
school = {Automation and Control Engineering},
abstract = {This thesis deals with the creation of an algorithm to detect anomalies or faults in a device. The algorithm input is current measurement and the output is the status of the device. The algorithm is obtained through machine learning techniques, in particular, support vector machines and feedforward neural networks are applied. The algorithm is obtained after a training phase in which data, collected from both faulty and healthy devices, are processed. Features are extracted from current measurement both in time and frequency domain. Also features that deal with the correlation between current and voltage measurements are proposed. The discussed procedure is applied to two case studies. The first one consists in the identification of noisy DC permanent-magnets gearmotors of two different models. Results, obtained both through support vector machines and neural networks, show that the algorithm is able to detect particularly noisy gearmotors, by using only time domain features. The second case study is about two dishwashers. The objective is to identify active components, in particular, if spry arms are moving and, in case, which one. The proposed algorithm for this task is obtained applying multi-class support vector machines through error-correcting output codes. The objective is achieved for one dishwasher while, for the other, the algorithm is not able to identify which spry arm is moving but can only detect if one of them is active.},
note = {Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Dr. Marco Lauricella},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
This thesis deals with the creation of an algorithm to detect anomalies or faults in a device. The algorithm input is current measurement and the output is the status of the device. The algorithm is obtained through machine learning techniques, in particular, support vector machines and feedforward neural networks are applied. The algorithm is obtained after a training phase in which data, collected from both faulty and healthy devices, are processed. Features are extracted from current measurement both in time and frequency domain. Also features that deal with the correlation between current and voltage measurements are proposed. The discussed procedure is applied to two case studies. The first one consists in the identification of noisy DC permanent-magnets gearmotors of two different models. Results, obtained both through support vector machines and neural networks, show that the algorithm is able to detect particularly noisy gearmotors, by using only time domain features. The second case study is about two dishwashers. The objective is to identify active components, in particular, if spry arms are moving and, in case, which one. The proposed algorithm for this task is obtained applying multi-class support vector machines through error-correcting output codes. The objective is achieved for one dishwasher while, for the other, the algorithm is not able to identify which spry arm is moving but can only detect if one of them is active.
3.
Cecchin, Leonardo
Graph-Based Exploration and Mapping Controller for Mobile Robots Masters Thesis
Automation and Control Engineering, Politecnico di Milano, 2020, (Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Danilo Saccani).
@mastersthesis{nokeye,
title = {Graph-Based Exploration and Mapping Controller for Mobile Robots},
author = {Leonardo Cecchin},
url = {http://hdl.handle.net/10589/167007
https://www.youtube.com/watch?v=on5oDsleTI0
},
year = {2020},
date = {2020-10-02},
urldate = {2020-10-02},
address = {Politecnico di Milano},
school = {Automation and Control Engineering},
abstract = {The focus of this thesis is the development of a novel controller for mobile robots used in exploration and mapping tasks.
The proposed solution is based on a reachability graph that represents reachable points of the environment and paths between them. It contains information about the topology of the surroundings, as well as the status of the exploration process.
The graph is updated as the robot moves, exploiting information from the sensors.
At the same time the graph is used to select the next target and to nd a feasible path to reach it. The proposed solution takes the name of Graph-Based Exploration and Mapping" (G-BEAM) controller. It is composed of two main parts: the obstacle avoidance controller and the exploration and mapping controller. The obstacle avoidance controller has the duty of guaranteeing that the robot does not collide with obstacles, like walls or objects. A novel algorithm has been proposed to that purpose, that is especially suitable for robots with fast dynamics. The exploration and mapping controller is in charge of the mapping and navigation tasks, updating the reachability graph with new information and exploiting stored data to control the robot. The controller is successfully tested through a simulation setup developed in MATLAB/Simulink environment.},
note = {Supervisor: Dr. Lorenzo Fagiano; Co-supervisor: Danilo Saccani},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
The focus of this thesis is the development of a novel controller for mobile robots used in exploration and mapping tasks.
The proposed solution is based on a reachability graph that represents reachable points of the environment and paths between them. It contains information about the topology of the surroundings, as well as the status of the exploration process.
The graph is updated as the robot moves, exploiting information from the sensors.
At the same time the graph is used to select the next target and to nd a feasible path to reach it. The proposed solution takes the name of Graph-Based Exploration and Mapping" (G-BEAM) controller. It is composed of two main parts: the obstacle avoidance controller and the exploration and mapping controller. The obstacle avoidance controller has the duty of guaranteeing that the robot does not collide with obstacles, like walls or objects. A novel algorithm has been proposed to that purpose, that is especially suitable for robots with fast dynamics. The exploration and mapping controller is in charge of the mapping and navigation tasks, updating the reachability graph with new information and exploiting stored data to control the robot. The controller is successfully tested through a simulation setup developed in MATLAB/Simulink environment.
The proposed solution is based on a reachability graph that represents reachable points of the environment and paths between them. It contains information about the topology of the surroundings, as well as the status of the exploration process.
The graph is updated as the robot moves, exploiting information from the sensors.
At the same time the graph is used to select the next target and to nd a feasible path to reach it. The proposed solution takes the name of Graph-Based Exploration and Mapping" (G-BEAM) controller. It is composed of two main parts: the obstacle avoidance controller and the exploration and mapping controller. The obstacle avoidance controller has the duty of guaranteeing that the robot does not collide with obstacles, like walls or objects. A novel algorithm has been proposed to that purpose, that is especially suitable for robots with fast dynamics. The exploration and mapping controller is in charge of the mapping and navigation tasks, updating the reachability graph with new information and exploiting stored data to control the robot. The controller is successfully tested through a simulation setup developed in MATLAB/Simulink environment.
4.
Trabattoni, Andrea
Design of a smart Ground-Winch for Systems of Tethered Multicopters Masters Thesis
Politecnico di Milano, 2017.
@mastersthesis{Trabattoni2017,
title = {Design of a smart Ground-Winch for Systems of Tethered Multicopters},
author = {Andrea Trabattoni},
year = {2017},
date = {2017-12-21},
urldate = {2017-12-21},
school = {Politecnico di Milano},
abstract = {In the last decade the interest in Umanned Aerial Vehicles (commonly known as drones) for civil applications has been constantly increasing across industry, academic research, governamental bodies. There is an incredibly wide set of tasks, such as eviromental mapping, precision agriculture, industrial infrastructure inspection, disaster relief. The main disadvantages of untethered UAV is the limited operational time that can be achieved before charging or swapping onboard batteries. To cope with this issue, tethered UAV has been developed in recent years. Within this thesis, the use of an aerial robot that is powered-over-tether by a groundwinch station is described. The idea is to simulate a system for remote powering small-scale UAV, aiming to achieve prolonged flight time in order to address possible civilian applications with such requirements. In order for the UAV to be capable of executing hovering trajectories as freely as possibile, a long power cable is wound on a custom-developed base, which is capable of releasing and retracting it when necessary. The base has been developed to operate in an autonomous way, locally sensing when the aerial vehicle requires additional cable length or when eccessive cable length requires retraction. The purpose of this workis to design the custom-device groundwinch mechanical layout, using components available on the market, and develop for it a proper control strategy. Moreover, an existing drone model and control system has been revised and improved to work with the groundwinch system. Simulations of the whole system have been performed: in order to give them the maximum reliability, parameters and data have been taken from the mechanical project. Finally a multidrone scenario has been proposed, supposing to have two multicopters tethered one to the other, in an innovative way with respect to the literature.},
keywords = {},
pubstate = {published},
tppubtype = {mastersthesis}
}
In the last decade the interest in Umanned Aerial Vehicles (commonly known as drones) for civil applications has been constantly increasing across industry, academic research, governamental bodies. There is an incredibly wide set of tasks, such as eviromental mapping, precision agriculture, industrial infrastructure inspection, disaster relief. The main disadvantages of untethered UAV is the limited operational time that can be achieved before charging or swapping onboard batteries. To cope with this issue, tethered UAV has been developed in recent years. Within this thesis, the use of an aerial robot that is powered-over-tether by a groundwinch station is described. The idea is to simulate a system for remote powering small-scale UAV, aiming to achieve prolonged flight time in order to address possible civilian applications with such requirements. In order for the UAV to be capable of executing hovering trajectories as freely as possibile, a long power cable is wound on a custom-developed base, which is capable of releasing and retracting it when necessary. The base has been developed to operate in an autonomous way, locally sensing when the aerial vehicle requires additional cable length or when eccessive cable length requires retraction. The purpose of this workis to design the custom-device groundwinch mechanical layout, using components available on the market, and develop for it a proper control strategy. Moreover, an existing drone model and control system has been revised and improved to work with the groundwinch system. Simulations of the whole system have been performed: in order to give them the maximum reliability, parameters and data have been taken from the mechanical project. Finally a multidrone scenario has been proposed, supposing to have two multicopters tethered one to the other, in an innovative way with respect to the literature.