The representation of the course of motion is an important task in many computer science applications: when, e.g., robots have to move autonomously through an open environment, locomotion of the robot and motion of other moving entities has to be tracked, represented, and processed. This is necessary for motion planning, for collision avoidance, and for interfacing with a human user or operator. A multitude of quantitative representations that solve such tasks are known. They can be successfully employed where sufficiently exact quantitative measurement data is given. When interfacing with a human operator, however, these quantitative representations can become problematic: Humans do not use to think in quantitative categories like "52 degrees", but in qualitative ones like "forward left". With naive users interacting with technical systems, it is important that communication with the system is as natural as possible to the user. Therefore, qualitative representations are especially important in the human-computer interface. One objective of the present work was to develop a qualitative representation of motion that is intuitive for humans and can serve for such human-computer interfaces. It is known that the Gestalt of motion is important in motion perception. Therefore, the Gestalt of a course of motion is a central issue in the developed representation. It consists of two layers: one is fine granular and uses qualitative motion vectors (QMVs) for motion representation. The second layer is coarser in granularity, more abstract, and uses the shape of parts of the trajectory as representational elements (SHAPE representation). The QMV representation can be generated from an observed course of motion; the SHAPE representation is created from the QMV representation through generalization, segmentation in basic SHAPES, and classification of sets of these basic SHAPES in complex SHAPES according to predefined SHAPE vocabularies. Different algorithms for performing the tasks of generalization and classification were developed. Considerations on reference systems for measurement and representation of motion, and on comparability and complexity of courses of motion close the theoretical part of the work. Some of the developed representations and algorithms could be applied in the navigation of a semi-autonomous wheelchair.     «

The representation of the course of motion is an important task in many computer science applications: when, e.g., robots have to move autonomously through an open environment, locomotion of the robot and motion of other moving entities has to be tracked, represented, and processed. This is necessary for motion planning, for collision avoidance, and for interfacing with a human user or operator. A multitude of quantitative representations that solve such tasks are known. They can be successfully...     »