Development of a Waste Collection Robot for Train Carriages Interior

In response to the increasing demand for quicker and less expensive train carriage cleaning, this project focuses on developing an autonomous robotic system for waste collection within train carriages. While numerous robots have been designed to clean the exteriors of trains, significantly fewer solutions exist for cleaning interior spaces. To address the unique challenges of detecting common waste items and reaching at the confined spaces of train interiors for collection, the paper proposes a novel proof-of-concept robotic system, designed to identify and collect waste items from the train floor surface, such as cans and plastic bottles from confined spaces, using an on-board waste collection and storage unit for proper disposal. The core focus of the project was to develop a waste detection algorithm for real-time object detection on low-power devices that navigate the robot, and to develop a waste collection mechanism that is durable and compact enough to be integrated onto a wheeled robot in order to gather the waste items. An NVIDIA Jetson Nano is used for detecting waste items using the SSD MobileNet v2 model (Single Shot MultiBox Detector) trained through the Jetson Inference project. The image-based waste detection is used to navigate the robot towards the waste item. A waste collection mechanism comprising of two parallel lines of revolving brushes is designed to collect waste and transport it to the onboard bin. This integrated system was evaluated to measure its performance in detecting and collecting waste items under a train seat. The results demonstrate the ability of this integrated system, to navigate towards the detected waste item and collect it from under the seat. Thus, showcasing its potential as an initial step towards creating an autonomous mobile robotic platform for waste collection in public transportation environments.