Research On Kalman Filtering Algorithm Based On DBN And Application In AGV

Automated Guided Vehicle(AGV)is a kind of automatic equipment equipped with automatic guidance device,under the premise that absence of manual guidance or driving,AGV can be carried out in accordance with the specified path,to carry out the loading and unloading of goods.Because of AGV's cargo handling,target traction and production line assembly and other areas of special purposes,as an intelligent robot,AGV becomes an essential tool in industrial production.Wheel movement is the main movement of AGV.AGV has simple structure,fast and efficient,good security and controllable advantages.In general,AGV system requires three important technologies: positioning,path planning and guidance.Positioning technology is the basic step of AGV navigation,is also the basic of follow-up tasks,only in accurately determine the current location and attitude,it is possible to find a driving path between the current location and target location according to the specific tasks.In this paper,we focus on the localization technology in AGV.In the commonly used AGV equipment,Kalman filtering(KF)algorithm is often used to fuse the information obtained on the sensor,and estimate the position and attitude of AGV.As the Kalman filter algorithm is an optimal estimation algorithm,it is easy to obtain good effect.However,when using the Kalman filter algorithm,it is necessary to pre-determine the statistical properties of noise,only in this way,KF can maintain accurate estimates.But,when the external environment is complex and changeable,the statistical properties of noise can easily change with the external influence,it is difficult to maintain accurate estimates of position and attitude by relying solely on the Kalman filter algorithm.In order to solve the above problems,this paper combines Deep Belief Networks(DBN)and Kalman filtering algorithms.By using the trained deep belief networks model to generate the adjustment factor,the covariance matrix of the noise is adjusted to maintain the Kalman filter algorithm for the AGV accurate estimation.Finally,the kinematic model of AGV is established for the structural model of actual AGV.The Kalman filtering algorithm and the Kalman filtering algorithm based on the deep belief networks are used to simulate and compare towards to localization problem of AGV.It can be seen from the analysis results that the Kalman filter algorithm based on the deep belief networks is more robust than the Kalman filter algorithm when the noise statistic characteristic is changed.The improved algorithm can maintain accurate estimation when positioning AGV.