Reinforcement Learning-Based Traffic Signal Control Using Speed Transition Matrix for State Estimation

The rapid development of Vehicle to Everything (V2X) communication and advancements in services profiled for mixed traffic flows are key drivers of solutions in the area of road efficacy and safety. Understanding the potential of the data provided by Connected Vehicles (CVs) obtained using V2X will enable the development of novel solutions to alleviate traffic congestion and improve safety. Obtained speed data from CVs can be used for Speed Transition Matrix (STM) generation, which is created by processing speeds of vehicles from two consecutive road segments. This research proposes a new intersection control system based on the STM and the application of Reinforcement Learning (RL). The developed intersection control system for the case of mixed traffic flows of road vehicles is tested in a simulation environment to confirm its effectiveness.