Enhance Recognition of Driver Profile by Improving OBD Protocol

Enhance Recognition of Driver Profile by Improving OBD Protocol

Bachelor Thesis, Master Thesis

An essential quality criterion of motor vehicles is their reliability. Reliability is equivalent to the likelihood that a product will operate faultlessly for a defined mileage under given operating and environmental conditions or under a given performance framework. In this context, the knowledge about the usage behavior and the resulting state of damage on public roads is a very important information. For example, this knowledge can be used to better design operating strategies and introduce predictive maintenance.

With regard to this problem, innovative solutions for load monitoring of automotive components are being investigates to predict the damage without installing any additional hardware. Previous works in this area have concluded that data available through the standard On-Board Diagnosis (OBD) interface of the vehicle can be used to predict the usage behavior and/or failure behavior of given components to some extent. However, the models and results are generally not as qualitative as the ones based on Controller Area Network (CAN) signals due to the enrich nature of the CAN bus. Indeed, the CAN bus is the internal communication bus of a vehicle allowing data transfer between control units. This means that more signals are available and that the signal quality is much higher. However, this bus is proprietary and not standardly accessible and interpretable in a vehicle. Being that, the objective of this work is to further improve models based on OBD data by investigating a way to enhance the OBD data or attempt to generate CAN data from OBD using generative models and use the generated CAN data to predict the damage.


  • Knowledge and Machine Learning and Neural Networks.
  • Experience in python programming language.
  • Familiar with Generative models.
  • (Bonus) Knowledge in keras and tensorflow.


  • Enhance the OBD data using generative model.
  • Attempt to generate CAN data from OBD using Generative models.
  • Use the generated data to train and build model.
  • Compare the results with model trained on OBD and another on CAN bus.

Please send your application to the following people:

  • Stéphane Foulard ()
  • Ousama Esbel ()
  • Alejandro Sanchez Guinea ()