High-speed Train Tracking Control Method Based On Integral Reinforcement Learning

The high-speed railway has the advantages of low energy consumption,high efficiency,and safe operation and has developed unprecedentedly in recent years.With the continuous advancement of intelligent control technology,high-speed train automatic driving technology has become a hot research topic in the industry and has been preliminarily applied,greatly improving the degree of automation and intelligence of train operation.With the increase of traffic density and running speed,the internal constraints and external environmental disturbances that affect the train operation are also increasing.Under the effect of time-varying dynamic characteristics of the train,the tracking control precision will be reduced and the control process will be unstable,which put forward higher requirements for the data perception and fusion processing ability of the train speed tracking control method.Therefore,through the analysis of the real-time running state information of the train,it is of great practical significance to study control methods that do not rely on the internal dynamics of the train system.Based on integral reinforcement learning control,this paper focuses on the study of train speed automatic tracking control methods.The main contents of this paper are as follows:Firstly,an adaptive tracking control method for high-speed trains based on integral reinforcement learning is proposed to solve the problem of reduced control accuracy caused by unknown train dynamics.Considering the existence of unknown parameters in the train model,a linear single-point-mass train model is established,based on which a state negative feedback control law is designed.In light of the unavailability of system dynamics,an adaptive control method based on integral reinforcement learning is proposed.This control method obtains the optimal tracking control strategy by carrying out the policy evaluation and policy improvement continuously and alternately,and adopts the adaptive control method to avoid involving the unknown system parameters.Simulation results show that under this control method,train can accurately track the desired position and velocity without involving the internal dynamics of the system.Secondly,a cruise control method for high-speed trains based on integral reinforcement learning and parameter identification is proposed to address the cruise control problem for high-speed trains with partially-unknown dynamics and unknown parameters.Considering the interaction forces between the connected vehicles,a multi-point-mass train model is constructed to describe the train movement process.Cruise phase has the characteristics of the longest running distance and fastest velocity,resulting in the increased external environmental disturbances.Thereby,a cruise control method based on integral reinforcement learning and parameter identification is proposed,which does not rely on the system internal dynamics and can achieve the asymptotic identification of unknown train model parameters.What's more,the simulation results verify the feasibility and effectiveness of the control method.Finally,considering the constraints of control input,a tracking control approach for high-speed trains with input constraints and partially-unknown dynamics is proposed to solve the unavailability problem of train dynamics caused by external disturbances.Considering the constrained train traction/braking force caused by actuator saturation,a novel cost function associated with time-varying input constraints is proposed.The nonlinear and uncertain external disturbances existing during the train operation are adverse to achieving the expected control performance.To address this issue,a tracking control approach considering the input constraints and partially-unknown dynamics is proposed.A neural network based on the actor-critic structure is utilized to solve the optimal control strategy.Simulation results show that the method not only achieves accurate velocity tracking control performance,but also keeps the control input within a predefined value.Besides,a high-speed train speed automatic tracking control simulation module is developed to illustrate the practical feasibility of the control method.