The recent trends in technology indicate Intelligent Transport System as an inevitable revolution sooner than expected. Recent years have witnessed active coordination between the industry and the research community to achieve the different automation levels. One of the critical aspects of an Intelligent Transport System involves efficient and accurate decision-making in real-time based on dynamic traffic conditions. The architecture should be such that the decision-making Machine Learning model’s performance should not affect the computation capabilities across various vehicular models. Correspondingly, a need for integrating a Machine Learning model that provides real-time decision-making while eliminating the need for large labeled datasets has arisen. In this paper, we propose an Intelligent Transport System architecture consisting of mobile and stationary Autonomous Vehicles and Software Defined Network (SDN) controllers. The Autonomous Vehicle acts as an Agent, and SDN controllers act as the edge devices. The Agent senses its dynamic environment for specific parameters, which are then sent to the edge nodes as input for real-time computation of decision-making model and returning the decision-policy to the Agent. The proposed architecture successfully eliminates the disparity in ML models’ performance, thus preventing accidents and traffic congestions. Additionally, we also propose a Reinforcement Learning-based decision-making model, which eliminates the need of collecting large labeled datasets thus reducing the overall cost for training the model.