Optimazation of Traffic Light Controlling with Reinforcement Learning COMP 3211 Final Project, Group 9, Fall 2017-2018

Ensuring the efficiency of transportation systems is a priority for modern society. Technological advances enable transportation system to collect huge sets of varied data on an unprecedented scale. Recent studies in combining deep neural network architectures with reinforcement learning techniques have demonstrated promising potential results in solving complex control problems with high dimensional state and action spaces. Inspired by these studies, we applied the similar idea to build a neural network architecture implemented by the reinforcement learning, where a new reward function for the traffic light control problem is implemented. Through experiment in SUMO traffic simulator, we prove that this approach could improve the efficiency of transportation system compared to the traditional controlling policy.