Research On Positioning Algorithm Of Autonomous Forklift In Indoor Operation Scene

Today’s manufacturing industry is changing rapidly from mechanization and electrification to informatization and high intelligence.Therefore,it is necessary to realize the autonomous transformation of construction machinery and operating vehicles.The first thing to solve for large-scale realization of indoor autonomous driving is the low cost positioning control of operating vehicles.In this dissertation,the position estimation of indoor operating vehicles and the positioning of operational targets are studied with insufficiently collectable information and poor performance of sensors.An autonomous forklift research platform was built for investigating the positioning and guidance in specific operating conditions.The algorithm for estimating the position of an autonomous forklift based on wheel speed information and UWB positioning information is investigated.Finally,an algorithm for estimating the position and attitude of the work target relative to the work vehicle based on visual information is proposed.Firstly,a basic software and hardware platform for autonomous forklift positioning and operation target positioning is built.The hardware platform contains a forklift test platform with autonomous driving capabilities,a main control computing platform,a developed vehicle positioning module and an operation target positioning module.The software platform consists of various modules for data storage,data sending and receiving,calibration,calculation and so on.The static and dynamic coordinate systems of the position estimation are set up in order to simplify the subsequent research process.Secondly,because of low update frequency of UWB positioning signal,poor signal quality and severe obstruction,a forklift position estimation algorithm based on the fusion of left and right wheel speed information and UWB information is proposed.The vehicle position is roughly predicted by the wheel angle calculation and forklift kinematics models based on the Ackerman steering principle.Then the extended Kalman filter and UWB information is used for getting preciser location of vehicles.The experimental results show that the maximum absolute error of the algorithm for estimating the vertical coordinate of the vehicle is 0.043 m,the average absolute error is 0.015 m,the maximum absolute error of estimating the horizontal coordinate of the vehicle is 0.057 m,and the average absolute error is 0.026 m.Finally,the target pose estimation algorithm for autonomous forklift operation based on visual information is investigated.To solve the problem of inaccurate recognition caused by susceptible interference from other finished products in the job scene,a marker-based job target recognition method is proposed.An adaptive ROI region segmentation method based on job target contour detection was designed though the fusion of color features and gradient features,then the visual imaging is analyzed,and the relative pose was estimated though the visual feature point information.The experimental results show that the maximum absolute error of the algorithm for estimating the vertical coordinate of the vehicle is 0.043 m,the average absolute error is 0.015 m,the maximum absolute error of estimating the horizontal coordinate of the vehicle is 0.057 m,and the average absolute error is 0.026 m.It performs accurately in automatic forklift control.