Research On Cooperative Positioning Of Security Robot With Multi-sensor In A Large Indoor Environment

Due to the advantages of all-weather and harsh environment operation,security robot is more and more used to carry out patrol and security tasks in public places.The positioning problem is the first problem to be solved during task running.The current positioning methods have the disadvantages of high cost,high hardware requirements,poor accuracy in indoor environment.In order to solve the above problems,the positioning method of security robot in indoor environment was studied,and a positioning method based on multi-sensor fusion was proposed to improve the positioning accuracy and efficiency.The security positioning experiment platform was designed,and the fusion positioning experiment was carried out to verify the positioning method.Firstly,the characteristics and positioning methods of domestic and foreign security robots studied,and decided to reduce the cost and increase the efficiency by improving the positioning method.The existing positioning methods,including conventional positioning methods and fusion positioning methods are studied,and the SINS,wheeled odometer and UWB fusion positioning are determined as the positioning methods of the security robot.Secondly,the research of UWB positioning is carried out.Based on error analysis and positioning demand analysis,DS-TWR and TOA are selected as ranging method and positioning method.The mathematical model of UWB positioning is established,and the influence of the number of stations,environmental noise and weighted algorithm on the positioning accuracy is studied.Finally,the weighted positioning of four stations is selected as the UWB positioning method.The platform of UWB positioning was designed,including the moving platform and the DWM1000 test module,and the TWR correction and UWB positioning experiment are carried out.Finally,the UWB positioning error is 0.306m,and the weighted positioning error is 0.265m,which reduced by 13.36%.Then,the research of dead reckoning positioning is carried out,including strapdown inertial navigation(SINS)positioning and wheeled odometer positioning.The dead reckoning positioning method is determined by defining the coordinate system and its transformation and the two-wheel differential model.The mathematical model of dead reckoning positioning is established.The positioning simulation of SINS and wheeled odometer is carried out,and the simulation results are used as the research basis of the fusion method.The dead reckoning positioning platform is designed.The data of SINS and wheel odometer were collected and the results of dead reckoning positioning are obtained through calculation.Subsequently,the fusion method is studied.The Bayesian filter and particle filter are studied,and the particle filter is adopted for fusion localization.The particle filter model is established,and the fusion positioning simulation was carried out by combining the UWB positioning model and dead reckoning positioning model.The location error is 0.080m,which is 68.25%lower than that of the single UWB location.The importance sampling and resampling processes in particle filtering are studied,whitch improve the positioning efficiency.The time of importance sampling process is reduced by 93.75%,and the running time of low variance resampling is reduced by 86.15%compared with polynomial resampling.Finally,the fusion positioning experiment is carried out.Firstly,the security robot fusion positioning platform is designed to collect the positioning information of SINS,wheeled odometer and UWB.Then upload it to the upper computer for fusion.Then positioning experiments are carried out to verify the positioning effect of fusion positioning method and improved particle filter positioning method.Among them,the average location error of the fusion positioning method is 0.176m,which is reduced by 23.14%compared with the single UWB positioning.The average positioning errors of polynomial resampling,residual resampling,stratified resampling and low variance resampling are 0.176m,0.170m,0.133m and 0.138m,which are reduced by 23.14%,25.76%,41.92%,39.74%compared with single UWB positioning.