Vehicles Routing in Dynamic Route Guidance System Based on Learning of Intelligent Agents

Nowadays one of the problems of big cities is increased number of vehicles for transportation of people and goods. Under this circumstances, congestion might be gently increased in urban transportation networks. Construction new roads and other urban transportation facilities are costly and time-consuming. On the other hand, the available urban transportation infrastructures cannot be used effectively, because drivers (especially private vehicles) are unfamiliar with various alternative routes and so they often choose longer ones. As a result, trips complexity is ever increasing. Therefore, one of the main challenges of traffic networks is guiding vehicles under dynamic traffic condition to their destination to reduce travel time using available network capacity more effectively. In order to reply to these problems, dynamic route guidance system (DRGS) is apparently an effective approach. It is among important sections of intelligent transportation systems (ITS). This system increases favorability of available infrastructures in transportation network and helps control of congestion on accessible time and space.The main core of dynamic route guidance system is computation of shortest path based on real-time information. This system can find the shortest path from starting node or zone to destination for drivers based on current situation and present to vehicles. Efficiency of this duty depends on shortest path algorithms. Also, using real-time information is possible by video or camera systems, magnetic loops and other traffic sensors on routes of transportation network.In transportation application especially in urban traffic networks, travel time from links depends on traffic volume, since static algorithms faces weakness in real-time conditions. Therefore, dynamic network condition must be examined for computation shortest path. Because of various approaches, solving dynamic network problems are more complex and Np-hard. In this kind of problems, choosing the optimized route is not easy but is a policy.Generally, with regard to necessity of research accomplishment, the general purpose of this research has defined in case of extension the robust routing strategy for route guidance systems under dynamic environment condition. In this paper the research methodology is combination of conceptual and quantitative models. At first vehicles route guidance in traffic networks remark to conceptual model, then the mechanism of working intelligent agents in network like quantitative model additive to solution problem. In this regard, this paper can be expressed as a route guidance conceptual framework based on decentralized routing structure by combining the e route guidance system and artificial intelligence. Thus we apply the concept of Agent-Oriented techniques emphasized on reinforcement learning (RL) as one of the alternatives toward overcoming uncertainty of vehicles routing in traffic networks. The main preference of reinforcement learning is ability of agent to learn for reaching a target, due to communicating control actions and affecting on the environment without exact model. In this regard we assume that some intelligent agents have been set in considered nodes and in fact, nodes of network become intelligent. These agents learn in addition to collecting information, and so they use previous experiences. Generally this research aims at traffic controlling through execution of an intelligent learning model on nodes of dynamic transportation network whose duty is guiding vehicles in route to target. The important results of this research are ability of learning models in presenting of route guidance strategy in adjustment to dynamic traffic condition and presentation of several route alteratives for drivers to reduce vehicle travel time criteria. Variable message signs usually inform drivers. This can be done by monitors inside the vehicles.In this paper the solution of research enquairy has been presented for grid networks and then for a small segment of Tehran traffic network as a real transportation network.