Research On Wheeled Mobile Robot Localization And Navigation In Social Environment

With the expansion of the mobile robot application scope as well as the rapid development of robot technology,people are increasingly demanding of the various functions of the mobile robot.In particularly,the accuracy,stability,security,and comfort of mobile robot localization and navigation in social environment is research focuses of mobile robot technology.In this paper,wheeled mobile robot in social environment is as the research object,and the key technologies that information acquisition and estimation of human body,the calibration of odometry systematic errors in localization method,localization and navigation based on multi-sensor information fusion,path planning based on social interaction space,and so on are studied.Its main contents are as follows:For static and dynamic human,the method based on histogram and bayesian statistics for estimating static human head orientation,and the method based on particle filter algorithm for estimating dynamic human motion information are presented respectively,so as to solve the problem that mobile robot in social environment need to acquire the human information(the information such as the body's position,posture,movement speed).For the estimation of static human head orientation,firstly,a square area is selected to represent the human head characteristics in the human head.Then,a kernel histogram based on kernel function is established.Finally,according to the information of the human head characteristics,the static human orientation is estimated by bayesian statistics.For the estimation of dynamic human motion information,the basic particle filtering algorithm(PF)and the extended kalman particle filtering algorithm(EPF)are used.Through experiments,the application and strengths and weaknesses of the basic particle filtering algorithm and the extended kalman particle filtering algorithm for the estimation of dynamic human motion information are compared.On the basis of analyzing the error sources of odometry system,the method of calibrating systematic parameters is proposed to reduce the systematic error and improve the localization accuracy.Firstly,the error sources of odometry system and its influence on robot motion are analyzed.Then,an experimental calibration method which consists of linear motion experiment and rotary motion experiment is proposed.The error is 295.042 mm when mobile robot finishes the 3 m × 3m square path movement after calibration.Namely the localization accuracy is decimeter level.In addition,the experiment method is simple,the experimental site is limited,and the experiment is easy to carry out(only need an about 4 m long,1.5 m wide rectangle area).The localization and navigation method based on multi-sensor information fusion is proposed under the condition that the motion path of mobile robot is determined.Firstly,the localization and navigation framework of mobile robot on the preset path is proposed.The sensor system of this framework consists of odometry system,electronic compass,camera and infrared sensor.Secondly,the adaptive extended kalman filter(AEKF)algorithm is used to fuse the odometry and the electronic compass data to improve the localization and navigation accuracy of the mobile robot.Before the use of data fusion,adaptive neural fuzzy inference system(ANFIS)is utilized to fit the pre-sampled data of electronic compass to improve the reliability of electronic compass data.In addition,a fuzzy algorithm is used to determine parameter k in the system noise covariance formula of AEKF algorithm to enhance the robustness of the AEKF fusion algorithm.Mobile robot path planning in social environment needs to consider the task constraints and social conventions,so a mobile robot path planning based on social interaction space is proposed.Firstly,model social interaction space based on task constraints and social conventions by using 2-dimensional asymmetric Gaussian function.Then,add the social interaction space model into the path planning,and the global optimal planning is carried out based on A* algorithm,so that the service robot can bypass the human body in a safe and comfortable posture.A mobile robot experiment platform has been established for experimental verification.Experiment is divided into two parts: the first part is to verify the feasibility and effectiveness of using adaptive extended kalman filter algorithm to fuse odometry and electronic compass data for localization and navigation through three preset path experiments in the indoor environment;The second part is to verify the feasibility of mobile robot path planning based on social interaction space through comprehensive scenario experiment.