Proposed Thesis


  • Cellular Network Dataset Analysis: Evaluate a dataset consisting of numerous traces taken by different vehicles, including bikes, cars, and trains. Each trace includes several fields, such as GPS position, received signal strength from nearby cellular towers (2G, 4G), and their GPS locations, among others. The recorded dataset is focused on Umbria. The goal is to analyze this dataset and create models for data inference. Many models can be used, such as regression, machine learning, and others. A good knowledge of the Python and C++ languages is required.
  • Smart Pot Application: Develop an Arduino-based application with the goal of creating an automated system capable of analyzing soil pH and moisture levels. The application will provide water to a plant pot if necessary. Machine learning can be utilized for providing feedback. Required skills include knowledge of Arduino, sensors, and the C programming language.
  • DJI Mini 3 Pro Flight API: Develop an API tailored to DJI SDK 5.x, focusing specifically on compatibility with the DJI Mini 3 Pro drone. While the official API does not support autonomous flight missions using the Wayline method with these drones, it does support the Virtual Stick method. Therefore, the Virtual Stick method needs to be adapted to mimic the functionality of the Wayline method. Proficiency in Android Studio and the Java programming language is necessary for this task.
  • Localization Mechanism using UWB PDoA Antennas: Develop a sensor localization system utilizing Ultra Wide Band (UWB) antennas, specifically leveraging the PDoA kit from DecaWave. The objective is to create a system that can be utilized by drones to localize ground sensors. This project involves both theoretical exploration and practical implementation tasks. Proficiency in Raspberry Pi, Linux commands, C programming language, and some hardware skills are necessary for successful execution.
  • Halyomorpha halys (HH) Lens App: Develop a cross-platform app for semi-automatic annotation of an image dataset within the bug detection system context. The app takes model predictions as input and should allow users to confirm correct/incorrect bounding boxes and verify accurate annotations. Additionally, it should enable the suggestion of bounding boxes missed by the model and facilitate the insertion and validation of annotations using a certain selection mechanism. Possible environments include Xamarin (C# or .NET) or Flutter (Java or Kotlin).
  • "Improve the comparison of weather data from established datasets with that collected by a custom microclimate station located within an orchard. This requires thorough analysis of diverse sources to assess bug detection rates in relation to current weather conditions and geographic location. Additionally, participate in configuring a Jetson Nano on the DJI Matrice 300 RTK to conduct machine learning-based recognition tasks as part of our continuous endeavors."


  • Preliminary comparison of weather data between Arpae dataset and a custom microclimate station located inside an orchard.


  • Implementation of drone-based delivery algorithms in a mixed Euclidean-Manhattan Grids.