Object Grasping Technology For The Mobile Manipulator In Indoor Environments

With the development of science and technology,robots play an increasingly important role in human's production and life.The mobile manipulators refers to the device that regards the mobile robots as the platform,on which mounts a manipulator on a mobile robot.The mobile manipulators not only has the environment perception and autonomous navigation function of the mobile robot,but also has the flexible operation capability of the manipulator.It is widely used in the fields of Industrial production,intelligent services,space work and investigation and detonation.This paper takes the reconnaissance and detonation of the mobile manipulator as the background,and focuses on the object grasping task of mobile manipulator,researching on the key technologies of the SLAM and navigation,object detection and grasping of the mobile manipulator in the indoor environment.The main work of this paper includes the following aspects:(1)Phantom X_ARM is built on Turtlebot 2 to form the mobile manipulator,and the RPLIDAR A2 and the Xtion Pro Live are equipped to build robot environment perception and object detection system.Besides,relevant software environment is configured.(2)SLAM and navigation of the mobile manipulator.In SLAM,the environment is perceived by the RPLIDAR A2,and the data is processed by Gmappimg algorithm.Besides,the number of particles,minimumScore and map_update_interval in the algorithm are analyzed by experiments,and the optimum performance parameters are obtained.In navigation,the RPLIDAR A2 is used to obtain the real-time environmental information,and the A* algorithm and AMCL algorithm are used to complete autonomous navigation of mobile manipulator.In the experiment,the navigation error is within 10 cm,which meets the design requirements and verifies the reliability of the algorithm.(3)Object detection of the mobile manipulator.Xtion Pro Live is used to obtain the real-time image information of the environment,and cv_bridge as the medium of ROS and TensorFlow communication,the object detection framework is built by using TensorFlow object detection API and the four Faster R-CNN ResNet of object detection algorithms are used to experiment.The experimental analysis shows that the target detection accuracy of Faster R-CNN ResNet50 algorithm reaches 88.29% under the condition of low CPU consumption rate,and the best detection algorithm is obtained.(4)Object detection of the mobile manipulator.Phantom X_ARM is configured by Movelt,and the object grasping experiment is carried out.finally,grab the target success rate by more than 85%,achieved design requirements.