Date of Publication

4-2025

Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Manufacturing Eng'g & Mgt w/ Specialization in Mechatronics & Robotics Eng'g

Subject Categories

Biomedical Engineering and Bioengineering

College

Gokongwei College of Engineering

Department/Unit

Manufacturing Engineering and Management

Thesis Advisor

Rhen Anjerome Bedruz

Defense Panel Chair

Armyn Sy

Defense Panel Member

Marlon Musngi
Ronnie Concepcion II

Abstract (English)

This study aims to demonstrate and investigate the use of a vision system in aiding a multi configurable robotic gripper in a pick-and-place operation for mangoes. Mangoes, one of the important fruits in the Philippines, are known for their various shapes and species that distinguish them from other fruits. Its asymmetrical profile can be a challenge for automated pick-and-place operations using machinery that may result in falls or bruising. Standard robot grippers utilize rigid end effectors that result in bruising on the surface of the fruit. The use of soft robotics such as finray may alleviate such concerns. Furthermore, to address the adaptability requirements of the asymmetrical profile of the mango fruit, the multi-configuration feature can be used to change the position of the finger of the gripper for added flexibility of the gripper during testing. A 3D printed fin ray gripper with multi-configurability was designed and simulated for the development of the vision-aided robotics system. The fin-ray utilized TPU filament in order to have a flexible material property while maintaining strength even with high elasticity. Different design parameters for finray design, such as edge thickness, number of crossbeams, and its angle were modified. Ridges were also added to add friction when grasping the mango. For the multi-configuration feature, three configuration modes were used for the experiment trials: parallel, t-shaped and thumb-3 finger. These modes were integrated with the vision system that utilized a CNN model trained using YOLO with an mAP of at least 90%. The experiment had two types: automated pick and place with orientation and then manual testing for further verification of configuration performance. From the automated trials, the pick-and-place system was able to grasp the mango types with the help of the vision system to change according to the mango type. At least 70% successful grasping rate was seen during the automated trials. Additionally, the automated pick and place showed that the vision system using a CNN was able to classify the mangoes most of the time and was able to determine the orientation using the bounding box of the mango. In the verification manual testing, it was reinforced that parallel configuration was good for carabao mango, t-shaped for pico while thumb-3 finger for katchamitha.

Abstract Format

html

Abstract (Filipino)

None

Abstract Format

html

Format

Electronic

Keywords

Computer vision

Upload Full Text

wf_yes

Embargo Period

4-13-2026

Download

Share

COinS