Trash Collecting Robotic System Using Two Autonomous, Mobile-Manipulator Robots with Convolutional Neural Network Object Detection System
Waste collection is an undesirable and hazardous task for workers. Autonomous robotics applied correctly to an effective system may lessen the demand for such work by humans. Further, intelligent robotics and the increasing access to visual processing tools make it possible for low-cost robots to incorporate convolutional neural networks in their control algorithms, and system of systems control strategies can be used to include multiple robots in a system. This thesis work aims to contribute documentation of a design process and a critical/constructive review of a functioning robotic system with two small, low-budget, tread-based manipulator robots relying on imaging and ultrasound sensors to collect empty water bottles in autonomously controlled harmony and transport such items to a collection area. While this prototype may not be fully equipped for use in an industrial environment, it will provide a demonstration of the available technology and its relatively underexploited application to an undesirable labor.