Josef Pauli Learning - based Robot Vision

visualization of the original scene Figure 4.9 shows an abstract depiction of the original scene. For the purpose of task decomposition we introduce five virtual points P1, P2, P3, P4, P5, which will serve as intermediate positions of the trajectories of the robot hand (see below). Generally, these positions are represented in the coordinate system of the robot arm which is attached at the static basis. Position P1 is the starting point of the robot hand for solving the task, positions P2 and P3 are located near the domestic area and serve as starting points for actively locating the target object and grasping the objects, respectively, position P4 is the point from which to start the servoing procedure for placing an obstacle object on the parking area, and finally position P5 is the point from which to start the servoing procedure for placing the target object on the inspection area. The decomposition of the high-level, deliberate task into sub-tasks, and the configuration and implementation of task-specific modules is based on an experimental designing phase. Therefore, in the following subsections the designing phase and the application phase are explained in coherence for each sub-task. 4.3.2 Localization of a target object in the image In the application phase the first goal is to find a target object among a set of objects which are located on the domestic area. Designing aspects for localization of a target object For the localization of a target object an operator is needed which should be robust and efficient (as has been studied in Chapters 2 and 3, extensively). Usually, the application phase leaves open certain degrees of freedom in arranging cameras and taking images. It is an issue of the designing phase to 4.3 Treatment of an exemplary high-level task 209 area parking domestic area inspection area CA2 CA 3 2