Estimation of congestion level at intersection points using AI

Congestion in vehicular scenarios has become one of the hot research areas among researchers. It is one of the most challenging issues, and it occurs when roads or channels become overloaded, mostly in highly dense network areas. Intersections on roads, generally called “congestion areas”, are places where most vehicles crash and accidents happen. So, controlling congestion is our primary motto. In this paper, the simulation over the map of Sonipat city (Haryana, India) is used via the Simulation of Urban MObility Simulator (SUMO). Various Machine Learning (ML) and Deep Learning (DL)-based models are used for calculating accuracy and calculating the R2 score. This study's findings show that gradient boosting offers the most promising approach for both congested and non-congested traffic conditions to real-time prediction of wait time. Using the gradient boosting model, an R2 score of 94.40% is achieved for the testing data. This paper provides an overview of various models for designing a strategy to avoid congestion-like situations for vehicular networks.