Derin Q Öğrenme Tekniği ile Trafik Işık Sinyalizasyonu

Creative Commons License

Tunç İ., Elmas Ö., Edem A. E., Köroğlu A. O., Akmeşe S. N., Söylemez M. T.

Otomatik Kontrol Ulusal Kongresi – TOK 2021, Van, Turkey, 2 - 04 September 2021, pp.1-6

  • Publication Type: Conference Paper / Full Text
  • City: Van
  • Country: Turkey
  • Page Numbers: pp.1-6
  • Istanbul Technical University Affiliated: Yes


Especially in big metropolises, problems such as inefficient use of time due to traffic congestion, air pollution and the effects of climate change are exposed. One of the main causes of climate change is traffic congestion, as it increases the amount of CO2 released. This problem negatively affects many living areas, especially metropolitan areas. In this study, traffic congestion problem is focused on an intersection. By using the data on the number of vehicles on the roads of the 4-way intersection as an input, a more efficient signaling method is obtained than traditional methods by training the traffic light agent with the deep Q learning method. In addition, it has been observed that performance of this alternative method used can be improved, and it is desired to reach the most appropriate value of the performance criteria by applying the input data to the deep learning model with different methods during the training process of the traffic light agent. It has been observed that learning with the Cell Method gives more effective results.