Title:Using AI for Mitigating the Impact of Network Delay in Cloud-based Intelligent Traffic Signal Control

Abstract:The recent advancements in cloud services, Internet of Things (IoT) and Cellular networks have made cloud computing an attractive option for intelligent traffic signal control (ITSC). Such a method significantly reduces the cost of cables, installation, number of devices used, and maintenance. ITSC systems based on cloud computing lower the cost of the ITSC systems and make it possible to scale the system by utilizing the existing powerful cloud platforms.
While such systems have significant potential, one of the critical problems that should be addressed is the network delay. It is well known that network delay in message propagation is hard to prevent, which could potentially degrade the performance of the system or even create safety issues for vehicles at intersections.
In this paper, we introduce a new traffic signal control algorithm based on reinforcement learning, which performs well even under severe network delay. The framework introduced in this paper can be helpful for all agent-based systems using remote computing resources where network delay could be a critical concern. Extensive simulation results obtained for different scenarios show the viability of the designed algorithm to cope with network delay.