Thornburg, D., Schmotzer, J., and Throop, M., "Vehicle Deep Data: A Case Study in Robust Scalable Data Collection," SAE Technical Paper 2017-01-1651, 2017, doi:10.4271/2017-01-1651.
Onboard, embedded cellular modems are enabling a range of new connectivity features in vehicles and rich, real-time data set transmissions from a vehicle’s internal network up to a cloud database are of particular interest. However, there is far too much information in a vehicle’s electrical state for every vehicle to upload all of its data in real-time. We are thus concerned with which data is uploaded and how that data is processed, structured, stored, and reported. Existing onboard data processing algorithms (e.g. for DTC detection) are hardcoded into critical vehicle firmware, limited in scope and cannot be reconfigured on the fly. Since many use cases for vehicle data analytics are still unknown, we require a system which is capable of efficiently processing and reporting vehicle deep data in real-time, such that data reporting can be switched on/off during normal vehicle operation, and that processing/reporting can be reconfigured remotely. We thus propose a distributed deep data collection topology that dynamically leverages the resources of individual vehicle modules, already present for delivering consumer features, in order to reduce hardware complexity, reduce overall system cost, and improve data robustness and reliability.