Font Size: a A A

Research On Urban Traffic Signal Control Based On Multi-Agent

Posted on:2009-08-15Degree:MasterType:Thesis
Country:ChinaCandidate:H X LiuFull Text:PDF
GTID:2132360242492892Subject:Traffic Information Engineering & Control
Abstract/Summary:PDF Full Text Request
With the continuous development of national economy, people's living standard improves gradually, vehicle population grows unceasingly, and urban traffic problem becomes obviously day by day. The traffic system is a complicated large scale system with randomicity and indefinability, The current signal control approach has obvious shortcomings in reliability of control and global optimizing and will unable to solve the increasing traffic problem ,make use of modern technology and intelligent method to solve urban traffic problem has attracted much attention of researchers.With the development of distributed artificial intelligence technology, Agent technology and Multi-agent system theory has provided a theory foundation for the study of software intelligent under distributed environment, as the urban traffic signal control is inherently distributed, it has very good prospect to adopt the multi-Agent technology to study urban traffic signal control problem.In this paper, combined with Agent technology, a distributed and coordinated traffic signal control system based on UTSCA is proposed. Firstly, this paper carried on structural design for UTSCA at single intersection, described its work process, and has carried on the design to the study pattern of the study unit, and then, carried on the structural design for the traffic signal control system based on UTSCA, and the formal description of the system. Then, selects traffic states of single intersection, forms quantitative description for the arrived cars and signal states of intersection through fuzzy classifying, and constructs the signal control rules, takes the total stop delay of cars for the optimized goal, uses the improved Q- learning algorithm to carry on the Agent training to improve the signal control rule. By comparing the method with traditional method through simulation, simulation result indicates that the effect of the new method in isolated intersection is better. Finally, carried on the description of the communication, coordinated way, and learning model in the signal control system based on Multi-agent, uses the distributed reinforcement learning method for coordination, the simulation result indicates that the method can reduce the average latency of the vehicles.
Keywords/Search Tags:Artificial Intelligence, Multi-Agent system, Urban traffic signal control, UTSCA, Distributed reinforcement learning, Multi-Agent coordination
PDF Full Text Request
Related items