Gianluca Palli

- Machine learning algorithms for robotic bimanual manipulation of deformable objects

This thesis is focused on the implementation of a robotic testbed for the development of machine learning algortihms devoted to the manipulation of deformable objects. The activity will be developed with the baxter dual-arm robot in the ROS environment.

- Design and implementation of Twisted String Acutation modules

- Ricerca e selezione di componenti ad alto contenuto tecnologico da applicare a macchine automatiche/utensili a fini diagnostici e predittivi

- Analisi, filtraggio ed elaborazione tramite algoritmi di machine learning di segnali provenienti da PLC a fini diagnostici e predittivi

- Analisi, filtraggio ed elaborazione tramite algoritmi di machine learning di segnali provenienti da macchine utensili CNC a fini diagnostici e predittivi;

- Analisi e implementazione di sistemi di visione per controllo qualità basati su algoritmi di machine learning nei processi di macchine automatiche

- Realizzare algoritmi di manutenzione predittiva basati su algoritmi di machine learning per caricatori di barre per torni

- Virtual commissioning di macchine automatiche per packaging 

- Standardizzazione e progettazione di una nuova HMI (Human Man Interface) per caricatori automatici di macchine a controllo numerico, basato su hardware SIEMENS e software TIA Portal (Totally Integrated Automation SIEMENS).

- Localization with Raw data from Laser

Currenlty the AGV's position is provided form the Sick NAV350 lasers. Objetive of the thesis is to develop an algorithm for the localization of the robot based on the laser scanner's raw data. The algorithm has to be implemented in a specific programming language and it will work with the current agv's architecture;

- Sensor fusion with 2 lasers

Objective of the thesis is to localize the robot by using two laser scanners. Each laser will provide information about position of the vehicle. The information must to be fused to obtain a more robust and reliable Agv's pose estimation;

- Traffic Coordination in Multi-AGV Systems

Developing of new strategies for coordinating AGVs to optimize the traffic in a warehouse. Topics:

• Deadlocks avoidance strategies
• Deadlocks solver
• Traffic Prediction
• Collision avoidance

- Advanced Anticollision

Analysis and developing a system for obstacle anti-collision. The project will be focused on set up and configuration of sensors and tools for avoiding the collisions with obstacles (protruding, moving, on the floor, etc.). Objective of the thesis is also to evaluate and compare of different sensors;

- Shape Profile Recognition

Analysis and developing of an application for identifying the shape of loading units. The application scenario is the identification of coils' shape. In particular, the system has to check if the coil is correctly wrapped and its shape complies with coil's requirements;

- 3D Map for natural navigation

Objective of the thesis is to analyze and to identify possible ways for building a 3D map of the environment. The map will be used for navigation, thus an algorithm to extrapolate useful information must be developed: first, 2D layer (2D map at specific height) can be extrapolated for navigation; second, useful 3d features has to be detected for 3d navigation;

- 3D Localization and Mapping

Developing algorithms for SLAM applications in industrial environments.

- Dynamic Home Positioning based on Traffic

Development of a method for identifying group of AGVs with respect to the congestion level in a specific area. This parameter will be used for setting up home positions in an automatic manner.