Erkennung und Behandlung von Ausnahmesituationen bei Robotern

Detection and recovery from exceptional situations is essential in robot-human cooperation in open and dynamically changing environments. We developed a modular, adaptable system for diagnosis and recovery from exceptions with clearly defined interfaces to planning and control components. Starting with the collection and characterisation of relevant exceptions and possible causes using an appropriate set of characteristics, this classification will be used as a basis for a systematic treatment of exceptions. In robotics, an operation-oriented view dominates, where interactions are composed of different tasks. Therefore, the basic idea was to transfer consistency-based techniques in diagnosis from a merely component oriented modelling to an operation oriented modelling as well. The recovery procedure defines the reaction to malfunction and strategies to delimit the effects of a failure and was done by a simple assignment for a start. The functions of the yet missing modules of the test bed for demonstration purposes showing the treatment of some typical examples had to be simulated