Task execution on heterogeneous robot platforms

The last decade has seen a significant increase in the amount and variety of robot platforms available to researchers, be it commercially available products or prototypes built by robotics researchers themselves. Due to a increasing demand for standardized sensors and actuators in industrial automation, most of these are composed of components that are comparable, if not the identical, and are similar in their structure. For the scope of this thesis, robots are mobile machines that are capable of perceiving their environment, are able to manipulate objects within their reach, and possess some computing capacity to control these abilities. Robotics researchers are nowadays able to present demonstrations showcasing these robots performing useful tasks in a specific field. These however are often lacking in terms of portability (how much work is necessary to have the system running on other robot platforms), generality (i. e. in how many other situations can the system be applied successfully), and robustness (how well can the system cope with mismatches between the estimated and the actual state of the world) due to focusing on the specific problem instance at hand. The goal of this thesis is to investigate methods to improve the former two properties for service robotics, by adding appropriate kinds of knowledge to the system. An assumption used throughout this thesis is that an abstract task description is available, which contains the formalized instructions defining the task.