Research On Learning Mechanism And Behavior Safety Evaluation Method Of Train Driver Agent

Recently, high-speed railway and the urban rail transit have been constructed, and then the importance of traffic safety has been obviously concerned and become hotspot. The cause of the driving accident is complex, including of the vehicle, lines, equipment factors and the human factors, such as the train driver, scheduling, management staff. Comprehensive analysis on important domestic and international driving accident, something will be found: because all sorts of equipment and security technology are improved and more perfect, the rate of content accident factors has been already reduced significantly, and the human factor is more and more obvious. In the operating process of the railway and urban rail transit system, human as the direct participator and management controller. Mistake is unable to avoid in the poor working condition and bad working conditions, so there will be accidents or potential risks. Therefore the human factor is impossible to eliminate. How to control and reduce human error is very important. The train driver is very outstanding as the direct train operator and controller.The main research of this paper is the train driver on urban rail traffic artificial system, taking the agent modeling method to construct the train driver agent and his learning mechanism. Its purpose is to make its dynamic learning ability. Through continuous learning, it will be able to produce actual "equivalent" behavior, and then take the evaluation. The main content of this paper is as follows:First of all, by collecting and analysis a lot of reference literature, it is summed up that the railway and the urban rail transit system. Present human research situation and the main research results have been put forward. The important of the human is pointed out for analysis. Scholars both at home and abroad mainly introduces the analysis of the train driver and the related research hot spot and achievements, involving selection rules of the train driver, physiological, psychological quality.Secondly, it is introduced that parallel control theory, artificial system and agent modeling method. Then the agent properties and categories are introduced in this paper. The learning algorithm for agent, and the comparison and analysis of their advantages and disadvantages of several algorithms are introduced.Thirdly, the train driver modeling method and the learning mechanism are established. From the study knowledge, learning level, learning mode and so on several aspects it is carried on that the detailed analysis and construction. The purpose is to make the train driver agent have dynamic learning ability. Also the implementation of the learning algorithm on the specific learning mechanism is designed, by selecting the Q learning algorithm, using the fuzzy comprehensive decision and fuzzy logic on the improvement.Finally, evaluation index is analyzed for the train driver agent' safety evaluation system, and using System Dynamics analysis method to establish safety evaluation system dynamics model. It is necessary to study the security evaluation of the train driver behavior after learning.