Research On Automatic Train Operation System Based On Genetic Algorithm

As the most efficient mean of passenger transport, urban rail transit has earned more and more attention. Urban rail transit construction of various cities of China is in fiery. The automatic train operation (ATO) system can substitute for the driver to control the train running. The rational designed ATO system can improve the efficiency and safety of train traffic, and can reduce energy consumption. This paper aims to design the ATO system which can control train to be safe, energy-saving, punctual, efficient and comfortable.Existing ATO control algorithm were mostly based on only one train traveling between two stations. On this situation that the interstation speed limit was all known. In this paper, there are multiple trains running between stations. It means the interstation speed limit is not all known. Thi s is the situation which is closer to the actual situation. This paper contrasted the control performance of a variety of ATO control strategies, and the optimal control strategy is:acceleration with maximum traction force, decelerating with maximum braking force, alternating coasting-traction when the train should be even running. The focus of the algorithm is that the determination of the switching points of coasting and traction. The proposed ATO algorithm is functionally divided into two layers:the searching layer uses Genetic Algorithm to search the optimal coasting starting point location and traction starting point location, search result is a position sequence of whitch the odd number of positon represent coasting starting point and the even number of positon represent traction starting point; protective layer predicts the operating condition of the train in accordance with the searched control state from search layer, and adjusts operating condition to prevent train speeding. In order to meet the requirements of different interval length and schedule time, the search layer using variable length chromosome genetic algorithm. In addition to the basic steps of the genetic algorithm, the adopted genetic algorithm includes two steps of process:gene duplication and gene deletion. In order to improve the fitness function, the paper proposes a novel performance indicator.In the searching process, the train model is need to predict train behavior, so by using the actual train running data, the train model coupling parameters are fitted(also use Genetic Algorithm). And then combining the train dynamic model, the train model is build up. Finally, the simulation system also uses the train model to evaluate the performance of various ATO control strategies. This proposed ATO control strategy is the optimal control strategy.