Gar-Bot: Garbage Collecting and Segregating Robot

Abstract

This paper presents the design and development of the first prototype of an automated garbage collection robot (Gar-Bot). It operates efficiently in an indoor environment. Main functionality of the robot is to detect and classify the garbage into 3 types, and collect them in different bins. To efficiently use the space in the bin, plastics are crushed after collection. Gar-Bot uses convolutional neural networks implemented on NVIDIA's Jetson Nano Developer Kit for garbage detection and classification. The IoT platform is used to notify the user with the status of the bin. The robot is equipped with a robotic manipulator, rotating bin for collecting garbage, Kinect V1, Processor and controller, and a base with wheels to facilitate navigation. The robot successfully detected and classified the garbage into various classes at an inference time of 5 seconds per frame, with an accuracy of 90 percent.