Research On Three Dimensional Mapping And Navigation Of Indoor Service Robot

Since the first remote manipulator was invented in 1947, the wide usage of industrial robot has greatly improved human productivity. And with the increasing maturity of robotics and intelligent technology, robot application scenario has also expanded from factory to many aspects of human life, in with the indoor service robot is one of the most important areas. Autonomous navigation is a core technology of indoor service robot researches, it's also the key and necessary condition of high-level services(such as security patrols and grasping object). The traditional solution of indoor mapping is using a two-dimensional laser sensor, and use the result two-dimensional grid map for navigation which is hard to meet the demand of navigation in complex indoor environment.In this thesis, starting from the demand of robot applications and key technology of autonomous navigation of indoor service robot, we have designed a low cost and efficient indoor mobile robot three-dimensional mapping and navigation platform. It is committed to to achieve a more accurate Indoor three-dimensional navigation scheme, and provide a consistent data interface for the high-level applications. The platform is based on the robot operating system(ROS) software framework, it has the capacity of the sensor data acquisition and processing, 3D mapping, localization and navigation, and some other complex applications. Based on the distributed node model and message pub-sub model, we have achieved dynamic node insert and data sharing between nodes and sending remote command. We used the Microsoft Kinect RGB-D sensor as the main sensor acquiring the 3D environment information, assisted by Hokuyo URG-04LX-UG01 two-dimensional laser sensor to improve the self-localization accuracy. Feature points matching algorithm and global closed-loop detection method have been used to ensure the accuracy of registration of 3D point cloud in the process of simultaneous localization and mapping(SLAM). We used a new data structure based on octree to storage 3D map data, it can reduce the data storage and calculation. By generating a 2D grid map projected from the 3D map, it can reduce the computational complexity of navigation and meet the demand of navigation in complex indoor environment.Finally, according to the design and platform, a series of tests and experiments using experimental robot were run to verify prove the availability and characters of the platform.