Constrained Markov Decision Processes for Intelligent Traffic

Traffic congestion has become a serious problem as most of the roads are busy and has to suffer vehicular traffic at an increasing rate. Proficient Signaling Systems are indispensable for reducing traffic congestion at intersections. This work tries to design a model according to constraint markov decision process for a given reinforcement learning problem statement which prediction of traffic congestion at the intersection and lanes in a busy roads. A Green Light Simulator is used to simulate real time data set of the Bangalore Roads and Traffic. After that several TLC algorithms are run and efficiency is computed manually. In the third step policy is build using parameters that contain features like acceleration, velocity, no of vehicles plying on the road. Based on these parameters, the policy generated helps to predict congestion at the busy roads. The dataset includes traffic data from Bangalore Roads.