Real Time Food Waste Segregation Using Robotic Vision

Loo, Jun Jie (2025) Real Time Food Waste Segregation Using Robotic Vision. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text Loo Jun Jie_FULL TEXT.pdf Restricted to Registered users only Download (13MB)

Abstract

This project aims to address the landfill issue and the problem of micro-plastics caused by unsorted food waste. This study explores real-time food waste segregation using robotic vision underscores the critical importance of addressing the significant amounts of unsegregated food waste that contribute to environmental pollution. While many advanced AI algorithms have been developed to tackle various challenges, there are relatively few autonomous robots specifically designed for food waste segregation tasks. The research has explored the fundamentals of autonomous robots and micro-controllers, revealing that autonomous robots are capable of making independent decisions based on sensory input and moving without human intervention. Microcontrollers play a key role in controlling these robots and ensuring they operate correctly. For this project, an autonomous food waste segregation system has been developed utilising robotic vision and a 6 Degree of Freedom (DOF) articulated robotic arm. The system integrates an AI-based food waste classification algorithm with a Raspberry Pi 5-controlled robotic mechanism. A parallel-jaw gripper is used as the end effector to autonomously pick and place non-biodegradable items from food waste. The vision system processes object coordinates and orientation to facilitate accurate manipulation, while a control system automates the entire process. The system’s performance is evaluated based on operational efficiency, including the time taken for processing, picking, and placing tasks. The total processing time is 4.053 seconds, while the times for picking and placing objects are 2.609 seconds and 2.404 seconds, respectively. Despite promising results, the system’s repeatability remains a challenge, indicating the need for further refinement in control algorithms and calibration. The success rate for picking and placing objects is currently 64%.