The Intersection of Reinforcement Learning and Traffic Light Scheduling

The goal of this project is to create a policy learner to optimize the traffic flow in a non-virtual city environment. However, in order to accomplish this task, we will be using model based reinforcement learning approaches to tackle the end goal. By first training the architecture on an easily testable and repeatable environment, we can develop a model that will complete this task quicker and easier than one reliant on real world training data. One key feature in this project is keeping the inputs simple enough that they would be readily available in real life applications - that is to say the features used as inputs should be easy to observe.