| KEYWORD |
DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA
AI-driven Picking Solutions for Industrial Feeding Machines and Applications
Thesis in external company
keywords ARTIFICIAL NEURAL NETWORKS, DEEP LEARNING, FACTORY AUTOMATION, OBJECT DETECTION, SMART ROBOTS
Reference persons ANDREA CALIMERA, ENRICO MACII, VALENTINO PELUSO
External reference persons GIAN LUCA DADONE, ALBERTO DALMASSO
Research Groups DAUIN - GR-06 - ELECTRONIC DESIGN AUTOMATION - EDA
Thesis type INDUSTRIAL R&D
Description The thesis project deals with the development of innovative AI-based solutions for a smart feeding machine. Specifically, the objective is to train a deep learning model for the regression of feature coordinates and optimal pick points to guide a robotic arm for pick-and-place industrial applications. The trained model will be deployed into the SupataŽ smart feeder by E.P.F., a robotic island driven by an AI-based vision system that handles components of various sizes, geometric shapes, and weights. The SupataŽ Smart Feeder consists of a vibrating table responsible for the singularization of objects and a robot arm capable of handling objects of varying sizes, shapes, and weights.
The objectives of the thesis include:
- Implementation of the training pipeline using a proprietary dataset.
- Application of some optimization and compression methods to minimize the latency of the inference engine, ensuring the required processing time on an industrial PC.
- Test and assessment of the inference model in an industrial environment.
Required skills The candidate is required to be proficient in Machine Learning/Deep Learning theory, Python programming, and training frameworks such as TensorFlow and PyTorch. Familiarity with optimization and compression methods for neural networks and experience with model deployment on embedded systems is a plus.
Notes - The thesis project is sponsored by E.P.F. Elettrotecnica S.r.l (www.epf.it), settled in Carrų (Cuneo, Italy).
- The work will be partially carried out in the company headquarters.
Deadline 31/12/2024
PROPONI LA TUA CANDIDATURA