Context-sensitive natural language generation for human readable event logs based on situation awareness in human-robot cooperation

Increasing complexity of robotic systems in respect to their abilities and reactions to environmental influences complicate the error and accident analysis. Even for experts the analysis of the vast amount of data gathered during robot operation is difficult due to the usually cryptic and unstructured form of data logs. In this work a system is introduced which analysis logs from a human-robot cooperation system and produces human readable event logs through natural language generation. For generation a template based approach is introduced which allows for adaption of detail granularity in order to facilitate different expert levels of end users. The information used during context-sensitive log generation is the result of different recognition and situation analysis methods. It is shown that the proposed system is capable of analyzing a data stream according to contextual relations and generates appropriate summarizations of occurred events.