Assessing the Accuracy of Vehicle Event Data Based on CAN Messages

Vehicles using controller area networks (CANs) for on-board device communications may have event data recorders (EDRs) that either capture or reflect network feeds from an array of sensors and other electronic control units. Using the data recorded in an EDR for investigative purposes requires external verification of accuracy. However, conducting external tests that set events can be expensive due to the time and equipment involved. This paper proposes a practical verification method that uses CAN bus monitoring tools to compare vehicle network traffic to external measurements. The premise of this work is that data reliability from an EDR can be determined if the reliability of the network data source for the EDR can be determined. Once the reliability of the source is determined, the reliability of the event data can be quantified based on effects of truncation and sampling. Furthermore, message timing and frequency can be quantified based on timestamps from the network analysis tool. Two examples of data analysis are given. The first example shows how to reverse engineer CAN messages from a Mini Cooper and the second example shows the comparison of J1939 message traffic to a sudden deceleration record from a Paccar PX-8 engine. Finally, a low cost hardware/software solution based on the Arduino platform is described to enable such measurements.