A Research On Vision Localization Module Of Educational Robot Based On ROS

With the rapid continuous development of robot technology,more and more industries have introduced robotics into their own fields,thus improving the productivity or solving specific problems in this field.Education robot is the product of the robot which is used in the field of education.It is an integrated system of intelligent techniques such as natural language understanding,speech recognition and synthesis,positioning and navigation,and seamlessly link up with educational environment,so as to provide learners with learning guidance based on scientific theories and reduce teachers' teaching burden.Education robot is an integration system of software and hardware,including the robot ontology of hardware and the software units to offer various educational services,of which localization and navigation technology is to solve the three questions,that is "where?" Where to go?How do you go?".Autonomous navigation is the basis of the education robot to provide other educational functions,and is a key feature of the intelligent equipment,which is different from the mobile phone and computer.Positioning is the prerequisite for navigation,while this technology is not mature,so positioning technology is also the research focus of this industry presently.The purpose of this paper is to design and implement a visual localization module,including the hardware module to support the visual-based location algorithm and the localization algorithm,which modularize the localization function of the mobile robot.The hardware module adopts embedded system based on Raspberry PI,and the RGB-D SLAM technology based on Kinect v1 is adopted for location.To sum up,the main works of this paper are as follows:Firstly,a ROS-based robot hardware module is designed and implemented to support the research of educational robot localization.Raspberry PI is used as a controller to integrate photoelectric encoder,ultrasonic radar,Kinect V1,servo motor,RFID and other sensors.As for the prototype,the PID algorithm is implemented to reduce motor error,the data acquisition interface of each sensor is designed,and the visual sensor Kinect is calibrated.Finally,the precision test of motion control is carried out,and the good control effect is obtained.Secondly,the SLAM technology based on feature points is studied,and RGB-D SLAM is selected as the research object of robot localization.The visual odometer part uses binary ORB features to extract the key frames from video streams,and then adopts FLANN to match the features from tow images.Finally,based on the motion equation and the observation equation based on the RGB-D camera model,a visual odometer experiment is carried out on the pre-collected data set,and the trajectory of the robot is estimated and the 3D point cloud map is constructed.Finally,the trajectory and map generated from the odometer are optimized by g2o,and the loop detection is added to make the 3D map closer to the real structure and the trajectory of the robot is more reasonable.Then the SLAM algorithm is implemented on the embedded robot platform,so the robot localization module based on embedded system is formed.