Smart traffic control of traffic lights based on traffic density in the multi-intersection network by using reinforcement learning

Smart traffic control of traffic lights based on traffic density in the multi-intersection network by using reinforcement learning


Smart traffic control of traffic lights based on traffic density in the multi-intersection network by using reinforcement learning

نوع: Type: thesis

مقطع: Segment: masters

عنوان: Title: Smart traffic control of traffic lights based on traffic density in the multi-intersection network by using reinforcement learning

ارائه دهنده: Provider: Seyedeh Monireh Mortazavi Azad

اساتید راهنما: Supervisors: Dr. Abbas Ramezani

اساتید مشاور: Advisory Professors:

اساتید ممتحن یا داور: Examining professors or referees: Dr. Ghaniei Zarch and Dr. Mansouri Zadeh

زمان و تاریخ ارائه: Time and date of presentation: 2023/02/14 16-17:30

مکان ارائه: Place of presentation: Seminar 1 electrical engineering department

چکیده: Abstract: Transportation and challenges have been always focused by researchers. One of challenges is controlling the traffic in urban area. Utilizing traffic light to prioritize vehicles passing by at the intersection is a solution to address the problems. In this thesis, smart traffic light was studied and simulated via SUMO. Two different models were considered. First of all, an individual intersection. The environment is discrete, and to take advantage of it, the state has more information, as a result, agent can work more efficiently. By using reinforcement learning, (Q learning and neural network), the traffic light is an intelligent decision at the level of the intersection to reduce vehicle consumption time by managing the allocation of phases. The results of this thesis showed that the length of waiting cars is reduced by applying the presented methods. In the second phase of this research, considering many previous assumptions, two intersections were investigated in a completely independent state and a state of sharing information between agents. The comparison between different definitions of reward and its effect on the volume of traffic has been investigated. Moreover, a study has been done on traffic monitoring through YOLO algorithm version ۵, which showed that the method of using this algorithm has impacts on the accuracy of identifying vehicles. Subsequently, estimating the volume of traffic. In this section, types of vehicles with respect to a coefficient, called PCU, were effective in creating traffic which makes it closer to the real conditions

فایل: ّFile: