Reinforcement Learning for Optimal Speed Limit Control Over Network
The goal is to optimize variable speed limit control (VSLC) strategies over network to improve both traffic safety and mobility. We propose to use graph-based deep reinforcement learning to improve the control effectiveness and scalability. The proposed research will advance the current knowledge and practice of VSLC in two aspects. First, this research will enlarge the scope of VSLC from link ¬based to network¬ based control to bring a new understanding about its system¬ level safety implications. Second, it will optimize the impact of VSLC using multi¬-objective learning approaches considering both safety and mobility.