| Service robot not only needs to improve the positioning accuracy in the indoor environment,but also needs to grasp the unknown objects in the environment.It is difficult to ensure the accuracy and reliability of the positioning and the object grasping only by the use of the vehicle laser or the vision sensor.Therefore,aimed at objects deliver applications of mobile robot in dynamic man-machine symbiosis environment,self-localization and grasping technology based on the combination of laser and RGB-D sensor is studied and the service robot mobile operation application software is developed in this dissertation.Firstly,in order to solve the problem of the classical particle filter localization method in dynamic indoor environment in which moving people are considered as the major interference,the rapid and reliable human body detection ability of RGB-D sensor was applied to segment the points cloud about detected humans,and then determined the range of the laser data that was caused by human interference when aligned with the laser coordinate to avoid detection of human leg from laser in the data.Based on the classical Monte Carlo localization,an improved particle filter algorithm based on the location matrix was designed.Using human detection results,the prior probability of the existence of unknown obstacles in laser data was estimated to solve the problem of the online estimation of the dynamic localizability matrix,improving the modeling ability of the dynamic localizability matrix to the unknown obstacles in the environment.The dynamic localizability matrix is applied to the dynamic correction of the particle set to ensure the reliability and robustness of the robot localization.Secondly,aiming at the problem of "eye out hand" position relationship among base-link,gripper-link and camera-link,a practical method for grasping the object of mobile manipulator is designed.The hand eye relationship of the mobile manipulator was calibrated and compensated by the combination of the vehicle laser sensor and the RGB-D sensor data.Aiming at the structural characteristics and grasping requirements of dexterous manipulator with two finger gripper,the top grasping strategy is adopted to simplify the description of the grasping posture of the end effector.In this paper,the methods of point cloud preprocessing were used:working plane were filtered,following by object segmentation and detection,geometric feature extraction and motion planning.The automatic grasping of simple regular shape objects was achieved at last.The typical function module of indoor mobile robot was developed on the basis of the above study,mainly including the positioning function program,eye hand calibration program and its program,object detection software and its testing procedures,grasping function software.Finally using the typical family environment as the application scenarios,experiments are conducted and the results confirm the method’s accuracy and reliability for robot localization and grasp. |
