THE FUZZY ATO WITH GA OPTIMISED COAST CONTROL

Abstract

This project is divided into two parts. The first part presents the development of the predictive fuzzy controller for the Automatic Train Operations (A TO). The A TO systems control the braking, powering of the train, coasting length, based on evaluation of riding comfort, punctuality, safety of the train and energy saving considerations. The second part regards to the development of the genetic algorithm (GA) that optimises the coast control of a train travelling between two stations. The optimisation is done before the train sets off to the next station. The fuzzy ATO controller has four groups of comprehensive rules that determine the most suitable status of the train for Motoring, Coasting, Brake To Target Speed or Brake To Stop. Fuzzy sets and indices based on the multiple objectives are pre-defined for the rules to arrive at the correct decision. Simulation results show that the proposed fuzzy ATO controller ensures that the safety limits of the train are maintained at all situations and achieves relatively good comfort. The controller also attains good punctuality and is successful in minimising mechanical braking. The drawback of the proposed fuzzy ATO controller is the lack of optimisation. This is the main motivation for the second part of the project to develop a GA based method for synthesising the train coast lookup table before departing from each station for an interstation run. The train's performance is improved further with the implementation of the coast control optimisation. Comparing the performance with the fuzzy ATO control, the ATO with coast control optimisation yields better performance based on evaluation of punctuality, energy saving and passenger comfort. The improvement is most significant under the circumstances of train congestion. The test results, although preliminary, have shown that the developed method is powerful, and has been applied to fulfil the multi-objectives control problem. The algorithm is reasonably fast and has potentials for online implementation.