Machine learning in traffic control to assist emergency vehicles on intersection transitions

Kurzfassung

Emergency Vehicles (like fire trucks and ambulances) played a vital role in the emergency situation. While emergency vehicle on the road, emergency vehicle’s driver must be drive with more concentration because of intersections on crossing road. In a real-world scenario, when a crisis arises, emergency vehicles cross the junction, the emergency vehicle driver must drive slowly and cautiously through every junction while using loud blue spotlights. During this time, a different road user may react suddenly, which might result in a collision with an emergency vehicle and another road users. A location might potentially be reached late due to traffic. This problem is important to emergency vehicle’s driver, other vehicles users and pedestrians. This master's thesis aims to investigate strategies for prioritizing emergency vehicles in a connected traffic system using a unique method from the topic of machine learning known as reinforcement learning. SUMO (Simulation of Urban Mobility), utilizing an existing machinelearning library and chosen traffic scenarios, is to be investigated using microscopic traffic simulation. As a result, we must maintain traffic flow while attempting to avoid obstacles. Here traffic signals help to direct the flow of traffic. Vehicle-to-infrastructure (V2I) communication is studying innovative approaches for building future-oriented solutions in traffic control. Prioritizing these in emergencies creates significant obstacles (another road users) to safe and effective traffic control. Using AI trained signal controller, the goal of signal controller will be to provide an intelligent method to prioritize emergency vehicles, allowing them to drive quickly, safely, and without any obstacles in an emergency situation. At the same time, it must have a minimum effect on other road users. With this aim, it will be possible to use modern connections and artificial intelligence technologies. The training results will be analysed using several reinforcement learning techniques to see if these AI control strategies can improve traffic behaviour. As a results, AI trained signal controller performed good results to improve traffic scenario at least 50% to 60% in emergency situation on an intersection.