Research On H_∞Filter Algorithm Based Train Integrated Positioning Method

Train Positioning Technology is one of the key technologies in the Automatic TrainControl (ATC) system. It ensures safe driving and high-efficiency operation and is also thepremise to have an effective control strategy over the ATC. The train integrated positioningtechnology is the trend for the development of the train positioning technology. Its highinnovation speed makes it necessary to reassess the situation and to find out suitable trainintegrated positioning system. When the external disturbance and model uncertainty are takeninto account, it is necessary to improve the robustness of the train integrated positioningsystem. Since the reliability of the train positioning information is more important than itsaccuracy, its reliability should be improved under the condition of its guaranteed accuracy atthe same time.The purpose of this thesis is to improve the reliability and precision of train positioninginformation from aspect as sensors, model, fusion algorithm and structure of the trainintegrated positioning system. The main contents of the thesis are as follows:(1) The Inertial Navigation System (INS)/Global Position System (GPS)/DopplerVelocity Sensor (DVS) combination instead of the classic INS/GPS combination is proposed,which is more precise and has a high fault tolerance ability.(2) The decentralized H_∞filtering algorithm instead of the traditional Kalman filteringalgorithm is proposed. Compared with Kalman filtering algorithm, the decentralized H_∞filtering algorithm is more effective in real train positioning situation, especially when theexternal disturbance is uncertain.(3) The random statistical model of train integrated positioning system instead of itsbasic model is proposed. Compared with the basic model, the random statistical model takesinto account the uncertainty of train integrated positioning system. It proves the superiority ofthe model by setting up the random statistical model of train integrated positioning systemand using the Linear Matrix Inequation (LMI) package to design the H_∞filter.In this thesis, The INS/GPS/DVS combination, loose coupling, and the federated filterstructure without feedback are adopted. On the basic model, decentralized H_∞filteringalgorithm and the traditional Kalman filtering algorithm are used separately to simulate. Andthe result of simulation shows that decentralized H_∞filtering algorithm improves robustnessof the train integrated positioning system. On the random statistical model, H_∞filter is used tofuse and the result is compared with that derived from the basic model. It shows that thereliability and precision of the train positioning information is improved by adopt randomstatistical model.