Adaptive Fuzzy Urban Traffic Flow Control Using a Cooperative Multi-Agent System based on Two Stage Fuzzy Clustering Fatemeh Daneshfar 1 , Javad RavanJamJah 1 , Fathollah Mansoori 2 , Hassan Bevrani 1 , Bahram Zahir Azami 1 1 Department of Electrical and Computer Engineering, University of Kurdistan, Sanandaj, Iran 2 Islamic Azad University of Buin-Zahra, Buin-Zahra, Iran daneshfar@ieee.org, j.ravanjamjah@ieee.org, fathollah.mansoori@gmail.com, bevrani@ieee.org, zahir@ieee.org Abstract— The traffic congestion problem in urban areas is worsening since traditional traffic signal control systems cannot provide efficient traffic control. Therefore, dynamic traffic signal control in Intelligent Transportation System (ITS) recently has received increasing attention. This study devises an adaptive and cooperative multi-agent fuzzy system for a decentralized traffic signal control. To achieve this we have worked on a model which has three levels of control. Every intersection is controlled by its own traffic situation, correlated intersections recommendations and a knowledge base which provides its traffic pattern. This study focused on utilizing the prediction mechanism of our architecture, it finds most correlated intersections based on a two stage fuzzy clustering algorithm which finds most intersections effect on a specific intersection based on clustering membership degree. We have also developed a NetLogo-based traffic simulator to serve as the agents’ world. Our approach is tested with traffic control of a large connected junctions and the result obtained is promising: The average delay time can be reduced by 42.76% compared to the conventional fixed sequence traffic signal and 28.77% compared to the vehicle actuated traffic control strategy. Keywords- MAS; Intelligent Transportation System; Fuzzy Control; Fuzzy Clustring; Traffic Light Control, I. INTRODUCTION Traffic congestion is a crucial problem in large cities. It is partially caused by improper control of traffic lights, which is not corresponding to the current traffic conditions. To alleviate traffic congestion in urban areas, the concept of Intelligent Transportation Systems (ITS) has been widely accepted in developed countries. ITS is a highly promising system for providing key solutions to current road congestion problems [1]. The problem of intelligent traffic control has been studied in the area of ITS for many years. We will refer to only a few which are related to our work. The first one is the method of Vehicle Actuated Signal Control (VA). This method controls traffic lights by considering the number of cars waiting in the queue to be serviced by a traffic light. When the current green light is going to be changed to red, but the sensor can detect that some cars have come in that range of distance, the duration of the green light is extended further [2, 3]. Some other methods are based on gathering traffic information during different times of day and year to help agents to make decisions [4, 5]. Using fuzzy control [6, 7], timed Petri Nets [8], SPSA [9], ant algorithm [10], knowledge based multi-agent system [11, 12], and a mobile agent [13] have also been suggested. The domain of traffic signal control is well suited for multi- agent based approaches owing to its distributed nature. Researchers and practitioners have now realized that single agent system [14], multi-agent systems and distributed artificial intelligence are attractive because they consider the social aspects of computer systems, ranging from human computer interaction over distributed problem solving, to the simulation of social systems [15]. But recently, multi-agent decentralized strategies for controlling urban traffic networks have attracted considerable attention. Roozemond and Rogier [16] proposed a prototype using agent technology to control traffic signal systems. Srinvasan and Choy [17] used advanced cooperative behaviors to improve individual agent’s learning process, Chang-Qing Cai and Zhao-sheng Yang [18] proposed an architecture composed of segment agents, crossing agents and section agents that it can realize an intelligent traffic management by sharing information. Ferreira et al. [19] also presented a multi- agent strategy where its agents optimize a traffic index based on its local state and sensors, and also on information from adjacent intersections. It appears that all these approaches lack a unified model and use only a fix number of neighbor intersections to predict traffic flow (the neighborhood area in most of them is limited with physical distance). Therefore we have proposed a model with a high abstraction, which considers most correlated intersections to predict traffic volume. Furthermore, each agent has the ability to communicate; enabling each agent to exchange relevant traffic information and cooperative with other multi-agents through a flexible architecture. The flexibility of this architecture is achieved through a modular design. To control the traffic volume in each intersection, we use three parameters: 1) Intersection Traffic volume (number of stopped cars behind red light and flow of cars from green light). 2) Correlated intersections traffic volume. 978-1-4244-2517-4/09/$20.00 ©2009 IEEE Author Personal Copy Author Personal Copy