Green-way traffic control pre-emption for emergency vehicles (GTPF-EV).

Abstract

Our project aimed at using a machine learning approach for traffic prioritization of emergency vehicles at traffic light junctions. The project also aimed at knowing the exit route of the emergency vehicle upon reaching the junction so as to efficiently utilize other routes. We trained an object detection model using 2,348 images where 1,400 images were training datasets, 801 images were validation datasets and 147 images were testing datasets. These datasets were obtained from available open source platforms and those taken from local roads around the university. The algorithm chosen was YOLOv4 due to its superior performance as compared to earlier version of object detection. Google Colab was also used to train our model on a free online GPU and result deployed on a local machine. The Arduino boards supported the design of our siren detector, EV route selector and traffic management system. Transceiver antennas were later integrated into these system in order to develop a bi-directional communication between the emergency vehicle and the traffic management system. The individual components built were later interconnected and tested as a single system.