Research On Localization And Tracking Method For NAO Robot Based On Multiple Sensor

In the current world trend of rapid technological change,the robot has gradually entered the personalized,intelligent era.There are so many sensors integrated into a humanoid robot,and shows the high technology level of modern intelligent robots,not only brings an abundant academic achievement for the scientific community,but also has made great contributions in anthropological sociology,participation in some meaningful projects,such as the rehabilitation of autistic children,treatment of depression,take care of “empty-nesters” and so on.In this thesis,all of research work will be accomplished on the NAO platform,which are based on the visual information and fused with other sensor information.Our work mainly includes object recognition,positioning and grasping,robot detection and tracking system design,robot trajectory estimation and walking error correction,software interface development.According to the physical characteristics of NAO and the method of image processing,two monocular ranging models based on the upper and lower cameras are established respectively,which can be used to quickly identify and locate the specified targets in the environment.Furthermore,in order to make NAO grasp the object after positioning,the coordinate system model of the robot whole body joint is established,and the forward and inverse kinematics of the camera,the left arm and the left leg to the center of torso are solved independently.In many cases,NAO needs to recognize the “partner” in the environment,while there are some limitations in the way of based on infrared equipment.So,we present a visual cognition-based method to complete the recognition of other NAO robots.In order to improve the tracking robustness,a robot detecting and tracking vision system is designed based on the fusion algorithm frame.In this system,the SVM + HOG feature extraction algorithm is the detection module,the Cam Shift algorithm is the normal tracking module,the EKF algorithm and the PF algorithm are the memory tracking module.Each module is fused with a certain logical relationship then to complete the entire system architecture.So the fusion strategy of the four algorithms can be generated adaptively by the system,but also can be selected by the user in the software interface.After NAO recognizes its own “partner”,they can perform the task together which we assigned to them.But there are have an unavoidable phenomenon that the robot will walk away from the pre-planned trajectory during this period.Therefore,according to the method of visual localization and tracking,the NAO Mark tag,ultrasonic,microphone and other sensor information are combined to correct the walking error,and assist the robot to complete the setting task.In addition,in order to estimate the motion trajectory of the robot in the course of executing the task,a monocular vision odometer is designed based on fusion of FAST feature and pyramid optical flow method,which is on the basis of robot step detection and statistical dead reckoning method,and the autonomous localization problem of the robot in the task environment is solved.Finally,we develop a PC software interface on the Visual Studio 2010 platform for NAO,which combines multithreading method,computer graphics knowledge,Open CV computer vision open source library and Tee Chart real-time curve drawing plugin and so on.