Autonomous navigation and mapping of mobile robots based on 2D/3D cameras combination

Presently, intelligent autonomous systems have to perform very interesting tasks due to trendy increases in support demands of human living. Autonomous systems have been used in various applications like houses, offices, museums as well as in factories. They are able to operate in several kinds of applications such as cleaning, household assistance, transportation, security, education and shop assistance
because they can be used to control the processing time, and to provide precise and reliable output. One research field of autonomous systems is mobile robot navigation and map generation. That means the mobile robot should work autonomously while generating a map, which the robot follows.
The main issue is that the mobile robot has to explore an unknown environment and to generate a three dimensional map of an unknown environment in case that there is not any further reference information. The mobile robot has to estimate its position and pose. It is required to find distinguishable objects. Therefore, the selected sensors and registered algorithms are significant. The sensors, which can provide both, depth as well as image data are still deficient. A new 3D sensor, namely the Photonic Mixer Device (PMD), generates a high rate output in real-time capturing the surrounding scenario as well as the depth and gray scale data. However, a higher quality of three dimension explorations requires details and textures of surfaces, which can be obtained from a high resolution CCD camera. This work hence presents the mobile robot exploration using the integration of CCD and PMD camera in order to create a three dimensional map.
In addition, a high performance algorithm for 3D mapping and pose estimation of the locomotion in real time, using the "Simultaneous Localization and Mapping" (SLAM) technique is proposed. The flawlessly mobile robot should also handle the tasks, such as the recognition of objects in its environment, in order to achieve various practical missions. Visual input from the CCD camera not only delivers high resolution texture data on depth volume, but is also used for object recognition. The “Iterative Closest Point” (ICP) algorithm is using two sets of points to find out the translation and rotation vector between two scans. Finally, the evaluation of the correspondences and the reconstruction of the map to resemble the real environment are included in this thesis.