Techniques to Synchronize and Align Driving Simulator Data
Kari Torkkola, Bob Leivian, Noel Massey, Chris Schreiner, Mike Gardner, John O. Summers · 2005
This paper describes techniques used to overcome issues encountered in collecting and aligning driving simulation data recorded from multiple devices. Driving simulators offer the ability to monitor the driver, vehicle, and environment in situation and in detail that are impracticable for roadway vehicles, but the data is often collected from various independent asynchronous data recording devices. Even when all subsystems properly collect data there exists a high degree of probability that the various data files will not be time aligned, thus resulting in misinterpretation of the data which then results in faulty conclusions about driving behavior or algorithms. The paper describes a data synchronization mechanism and a multistage alignment process whereby data from various subsystems is aligned in a two step process: rough alignment and then optimal alignment within milliseconds. Optimal alignment uses a variation of the dynamic time warping (DTW) algorithm resulting in a single data file (plus video). The paper also presents the results of the alignment process on driving simulator data and also describes an automated data validation technique to ensure the integrity of collected data.