Multi-agent and Artificial Neural Network for Traffic Lighting Optimization

Agent technologies are becoming more and more common in distributed systems, but their potential for sophisticated automobile traffic control has not been fully investigated. In this paper, an artificial neural network and agent technology-based traffic simulation system is presented. By utilizing both intra—and inter-intersection collaboration and coordination, this framework seeks to optimize amount and the traffics at isolating intersections as well as for the entire multi-intersection network by acting on phase layout indicated by the sequence and lengths. Real-time optimization of signal design is carried out, taking into account both local stream factors and the traffic stream circumstances at nearby intersections. In order to break down traffic controlling optimization to smaller problems and facilitate distribution resolutions, the system leverages agent cooperation, communication, and coordination capabilities in addition to decentralized organization. Thus, parallel tasking allows for the quick resolution of the individual components. Additionally, it makes use of artificial neural network technology to manage traffic condition unpredictability. The suggested framework was designed and validated in an instance. Analysis and instantiation results show that considered method can greatly increase efficacy both at a single connection and over the network of multiple intersections. In comparison to other schemes, it lowers the regular traveling delay and amount of time spent.