Architecture of on-line data acquisition system for car on-board diagnostics

Abstract

Modern cars produced for the last two decades are full of electronic devices called Electronic Control Units (ECU). They are responsible for collecting diagnostic data from different components such as the engine, breaks etc. using probes and sensors. The collected data are validated against built-in heuristic and abnormal behaviour is reported to a driver by a gauge on an instrument cluster. ECUs use data provided by other ECUs. Information is transmitted over the dedicated network called Controlled Area Network (CAN). Every car equipped with ECUs and CAN exposes information over universal diagnostic interface called On-Board Diagnostic. Using the interface, it is possible to gather car's live data. With the data mining approach, it is possible to exploit the collected more effectively to obtain much more information about the functioning of car components than it is provided by standard vehicle equipment. The paper describes how to build a laboratory set to facilitate automated data collection. It consists of three major components: data acquisition, automated logs collection and persistent storage with presentation tools. The first component is based on Torque application for which reverse engineering was performed.