A Cross-Platform Data Acquisition and Transformation Approach for Whole-Systems Experimentation – Status and Challenges

摘要

The emerging availability of data acquisition, processing and analysis tools as well as multi-modal data sets originating from long-term robotics experiments facilitate collaborative and comparative research on Human-Robot Interaction (HRI) at the systems level. A key requirement for sharing and using experimental data within the HRI domain (but also beyond) is openness. Infrastructure for recording and processing experimental data requires standardized protocols and formalization of the syntax and semantics of the underlying data formats. While these aspects can be specified explicitly for a single system, experiment or associated framework, the development of a more general approach for this problem would be beneficial. Such an approach would not only allow to record data from different sources, e.g., the various robot hardand software platforms but also external sensors for recording ground truth such as eye or motion tracking devices typically used in HRI research. Furthermore, the integration and data processing with various analysis tools (such as Matlab, R, or annotation tools such as ELAN [6]) would become much easier in this case. In the following, we will briefly describe the current state of a data recording, transformation and processing approach that explicitly targets openness and highlight particular challenges related to this goal.