Hawkmail
Simulators & Vehicles
Research
Media & Events
Staff
| Document Number: | N2005-026 |
| Document Type: | Conference paper |
| Author(s): | Sunil Bulusu Yiannis Papelis Bin Chen Matt Schikore Omar Ahmad |
| Publication / Venue Name: | DSCNA 2005 |
| Publication Date: | 2005-11-01 |
| Abstract: | A key problem with data collected during experimental studies is the expeditious reduction and delivery of reduced data. It is not unusual to generate over 500 MB of raw data for a 20-minute drive on the National Advanced Driving Simulator. Managing, transporting, and reducing the simulator data is a time-consuming process that traditionally starts at the end of data collection. During pilot studies, which are conducted to test the feasibility of an experimental design, speedy reduction of data and prompt statistical analysis is a crucial step in designing the full experiment. In such cases, the time-consuming process of reducing large data files impedes progress and escalates the cost of studies. In order to meet the cost and time requirement of studies, an alternative concept of embedded data reduction is used. Embedded data reduction uses scenario authoring elements that embed information in the raw data stream to calculate performance variables. During experimental runs, this information is permanently embedded in the large data files collected. When importing the data files, performance measures embedded in the data streams are recognized and interpreted, and the reduced data is generated. This paper covers the design and analysis of NADS Embedded Data Reduction and its effect on the experimental design process and scenario authoring. In addition, the paper also includes a brief comparison to traditional post-processing techniques and some of the downsides of embedded data processing. |
| Body: |  |
| Copyright: | DSCNA |
| Keywords: | None listed |
