Diacon, A., Daily, J., Ruth, R., and Mueller, C., "Accuracy and Characteristics of 2012 Honda Event Data Recorders from Real-Time Replay of Controller Area Network (CAN) Traffic," SAE Int. J. Trans. Safety 1(2):399-419, 2013, doi:10.4271/2013-01-1264.
Prior EDR testing methodologies required setting events in the airbag control module in the vehicle during controlled driving behavior. Duplicating events was nearly impossible, and it was difficult to separate how much differences in recorded speeds to reference speeds was due to measurement error, wheel slip, reporting time delays, or data truncation within the EDR. Recording thresholds have also increased making non-deployment and deployment events closer in magnitude, increasing the risk of accidentally exceeding the deployment threshold while setting events. The new methodology eliminates the risk of accidentally deploying airbags while gathering GPS and CAN bus data in the test vehicle. The techniques presented in this paper also allows gathering of data in vehicle without tampering with the airbag control module, which reduces the potential liability to testers using rental or borrowed test vehicles. The new methodology allows for repeatable testing and mapping the transfer function between the vehicle CAN bus data and the EDR. Should a manufacturer make a design change to an ACM EDR, identical inputs can be given to exemplar ACM's from before and after the changes to document any change in the transfer function. This methodology allows researchers the ability to re-create events of interest in a low-cost, repeatable manner.The accuracy of the 2012 Honda CR-V and 2012 Honda Civic event data recorders were tested using this new two part methodology.First, the test vehicles were instrumented with both a Racelogic VBOX differential GPS speed measurement system and a Vector CAN data logger. The measurements from the VBOX were transmitted onto the vehicle's Controller Area Network (CAN) bus that also contained messages reflecting indicated vehicle speed, brake status, accelerator pedal position, steering wheel angle, individual wheel speeds and other signals. This put the GPS speed data on the same time base as the vehicle CAN speed signal such that no additional synchronization was required. This permitted analysis of the accuracy and update rate of the vehicle speed CAN signal, which is the source for speed data used in the Event Data Recorder (EDR).Second, a system was developed to replay the recorded CAN data to an exemplar airbag control module in the laboratory, such that the exemplar was receiving data exactly as if it were in a moving vehicle. A pneumatic fixture with a slide was built to allow the exemplar module to be accelerated to nearly 10 km/h (6 mph) and then stopped in about 80 msec. to create a non-deployment event that met the minimum 5 mph delta-V over 150ms threshold. Actuation of the event setting fixture was computer controlled (using LabVIEW) and synchronized with the CAN replay system so that the desired test condition could be replicated precisely. An external accelerometer was mounted to the airbag control module to mark when the ACM began recording and the accelerometer data was matched to the acceleration and delta-V data in the EDR. The desired test conditions were replayed to the airbag control module and a series of non-deployment events were set. Each event on the EDR data was read using the Bosch Crash Data Retrieval system.The EDR data was compared to the inputs, and it was determined that the two byte vehicle CAN bus signal was truncated to the next lower whole km/h when recorded in the EDR. Under steady state conditions the speed data was accurate within 2%. The vehicle CAN signal published new values every 0.1 seconds, and the CR-V updated values every 0.1 seconds, but the Civic only updated every 0.6 seconds during hard brake events. During hard ABS braking events the expected wheel slip led to the vehicle speed signal under-reporting ground speed, but the delay in the CAN bus update rate resulted in over-reporting ground speed in many of the tests.