polito.it
Politecnico di Torino (logo)

An artificial intelligence system to improve the performance of anthropomorphic robots in the apple packing process

Damiano Scibilia

An artificial intelligence system to improve the performance of anthropomorphic robots in the apple packing process.

Rel. Alessandro Rizzo. Politecnico di Torino, Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica), 2021

Abstract:

Despite of the technological advancement that mankind is experiencing, there are numerous companies where some workers are still employed for alienating, repetitive and unsatisfying tasks. In the hope of giving these people a new, more rewarding job prospect, a software program has been developed to automate the process of apple packing. The state of the art in industrial automation in this context is limited to the use of an additional secondary roller, which picks up the fruit individually in order to make it easier to hook. The current project, on the other hand, is innovative in several respects: firstly, it aims at not using additional mechanical supports besides the robot itself, thus increasing the adaptability of the robotic system. In addition, the goal is to observe also the color of the apples, in order to position them using also this parameter. Finally, it is intended to use cheaper sensors than those previously employed in other similar systems. First of all a feasibility study has been carried out by means of a simulator that has allowed to identify the configurations that accept a robotic system of this type and in particular it has been understood that to meet or exceed in speed a human operator more robotic stations are required. Then the vision part was followed: apples and tray cavities were searched in the images first by means of pattern recognition tools and then by neural networks. The neural networks gave very satisfactory results and were therefore used in the final software. The color of the apples has been detected by methods that include the use of neural networks and also by scripts that do not involve their use. The result obtained demonstrates how scripts that evaluate apple color are far faster and more effective in accomplishing the task. The innovative use of neural networks in this project can not be left behind in future developments given their adaptability and reliability. Further work on this project must therefore address the detection of apples in space and their actual placement on the trays.

Relatori: Alessandro Rizzo
Anno accademico: 2021/22
Tipo di pubblicazione: Elettronica
Numero di pagine: 122
Informazioni aggiuntive: Tesi secretata. Fulltext non presente
Soggetti:
Corso di laurea: Corso di laurea magistrale in Mechatronic Engineering (Ingegneria Meccatronica)
Classe di laurea: Nuovo ordinamento > Laurea magistrale > LM-25 - INGEGNERIA DELL'AUTOMAZIONE
Aziende collaboratrici: E.P.F. Elettrotecnica Srl
URI: http://webthesis.biblio.polito.it/id/eprint/20450
Modifica (riservato agli operatori) Modifica (riservato agli operatori)