The ever-growing demand for smarter cities has paved the path for envisioning smart traffic, in the existing road infrastructure. Learning-based techniques have recently been promising for solving traffic management problems in Intelligent Traffic Light Systems. This paper proposes a novel Intelligent Traffic Light System for the roundabouts accounting high density traffic scenarios. The roundabouts have shown excellent results in terms of reducing traffic and accidents. The proposed agent is tested using two algorithms, viz., Deep Q Learning, Actor Critic Agent-based Reinforcement Learning (RL) along with an effective reward computation and state functions. The results of the proposed algorithms are compared and analyzed for the most suitable learning technique in the traffic scenario.