Distributed Management and Representation of Data and Context in Robotic Applications

The traditional, isolated data handling in sensor-actuator systems does not fulfill the requirements of robots that need to interact with their smart environment. Consequently, we have to develop new mechanisms for adaptive data and context handling. We firstly investigate what types of data are present within smart environments and how they can be classified and organized. Only if the available data can be structured, it can be queried and thus put into context. This is important because the variety of data and possible interpretations is tremendous, ranging from measurement values, sensor and robot descriptions/states/commands, to environmental data, such as positions, maps, spatial relations, etc. To cope with this diversity, we developed a solution capable of storing and accessing data within a distributed environment by providing additional context information. Furthermore, we describe how this information can be assembled in a task-oriented manner. This enables robots to dynamically generate environmental abstractions by using data from different sources and also enables them to incorporate external sensor measurements.