Overview / Usage

Having identified the challenges of traffic in Hong Kong, it is urgent to provide alternative solutions to optimize the traffic condition. The proposed solution aims to address the problems with deep reinforcement learning, which only requires minimum hardware setup on the traffic light instead of vehicles. The whole system is an IoT network (Azure IoT Suite) with multiple sub-systems installed on each traffic light and a sustainable backend server deployed on Azure Cloud Services. For each sub-system, there is a chip for controlling the traffic signals, a camera for observing the current traffic condition, a set of Windows 10 IoT core which is in charge of coordination and communication within the sub-system. Given these configurations, the IoT network is able to facilitate traffic flow by implementing reinforcement learning and vehicle counts with computer vision (powered by Azure Cognitive Service). The workflow, algorithm, and platform of the project will be illustrated in the following paragraph.