Data Processing Framework Supporting Large-Scale Driving Data Analysis

The increasing number and complexity of advanced driving assistance systems require an ever-increasing amount of field data. Data corresponding to the actual use of vehicles by ordinary drivers in real driving conditions is needed to calibrate new systems, identify their shortcomings, and evaluate their impact. Although necessary, experiments in controlled conditions, such as test-bench, test-track, or driving simulator, only allow observations in a finite number of predefined and generally simplified scenarios. In this session, we show how we have implemented, using MATLAB, a very powerful framework that automates data management and processing and completely separates technical tasks from more scientific ones. Using this software, analysts have all the necessary tools to browse and search for specific things in data, transform the logged records into usable results with their own algorithms, and visualize and annotate everything (original signals, derived data, geographical position, video, etc.) in a single interface.