A quantum Lévy flight optimization based peripheral Yolov5 vison for scaling and optimizing autonomous train operations during transmission

Abstract

In Autonomous Train (AT) operations, predicting future movements of trains to ensure the safety and efficiency of this public transport, is a critical task performed in the yard and terminal stations or between the source station/terminal and destination. Artificial Intelligence (AI) enabled Autonomous Vehicles (AVs) analyse their surroundings and make informed decisions based on the collected information. AVs learn their surroundings through perceptual motion, and predict the movements of various agents, such as humans, bicycles, and animals, through the prediction phase. If the predictions are accurate, AVs take proactive measures, making travel safer for everyone. However, the uncertainty in predicting agent/object’s behaviour remains substantially difficult, despite significant research being conducted in this area. The objective of this study is to investigate the optimisation of AT Operations during Transmission (ATOT), with the goal of reducing wait times and ensuring safety even in the presence of obstacles, particularly animals. Quantum Lévy flight optimisation (QLFO) addresses these concerns on operational scalability of ATs. It generates real-time spatial data to make these operations efficient and to reduce the average waiting times. The YoLoV5 object recognition identifies the potential risks, such as individuals traversing rails or animals lying along the way and either rerouting or halting ATs to prevent crashes. The QLFO detects problems either before or after the train arrives at a junction or terminus and takes proactive measures using YoLov5 evaluation on the satellite image captured over the rail track. Evaluation experiments show that the proposed ATOT achieves a 9.3% reduction in the average waiting time spent.