Low-Cost Autonomous Trains and Safety Systems Implementation, using Computer Vision

Abstract

The need of modern transport solutions is a tendency that has been developed also in the railway transport. This study provides a possible implementation of a fully autonomous train system with low impact on the railway infrastructure, using computer vision and machine learning concepts. It could be implemented on various existing safety and infrastructure systems. The system has been tested on a H0 scale modified model train and a Raspberry Pi with a Pi Camera as processing unit. The proposed system combines several software and hardware technologies into a single embedded system that provide the required safety on railways and can set the trend for real trains. Furthermore, the main motivation of the concept is that the railway transport automation represents an essential step in transforming this domain into one as flexible as road transport. In this regard, over the years, a multitude of control and safety assurance systems, based on various technologies have been developed to lead to the most optimal outcome. The primary innovation of the study resides in the application of neural network quantization to enhance temporal efficiency, alongside the advancement of a comprehensive autonomous railway transportation system.