Data collection framework for vehicular On-Board-Diagnostic systems

Abstract

Most modern vehicles contain an On-Board-Diagnostic (OBD) system that can collect a wide range of system data from the vehicle. In aggregation, such data could be applied towards solving the problems of accident prevention, vehicular safety, and driving efficiency. In this paper, we describe our design and implementation of a distributed vehicular data collection framework for running applications that aggregate customized OBD data on demand. Our framework is low-capital, low-maintenance, and flexible enough to handle several types of data requests. In our evaluation, the framework achieves within 2% accuracy for data collection at 1-second interval when compared to an externally calculated benchmark. We also simulate data collection under non-ideal conditions and methodically characterize the drift of accuracy as total percentage of packet loss rises for different patterns of data loss; in our experiment, we conclude that for our sampled data, dropped blocks have the greatest impact on accuracy..