Big Automotive Data Analytics: SEnsor Modeling and Performance Analysis (BADA-SEMPA)

Purpose and goal

The analysis of sensor data plays a crucial role to build highly automated and autonomous vehicles. Such analysis makes it possible to develop better sensor verification and accurate computer-aided engineering (CAE) simulations, and to better implement the active safety functions. The main research questions addressed by the project were related to improving ground truth, developing sensor models, and semi-, un-supervised data analysis. During the project, we have achieved the main goals and developed different analysis methods to improve the verification of AD.

Expected results and effects

Findings are presented and discussed in different conferences, e.g.: 1- J. Florbäck, L. Tornberg, N. Mohammadiha, “Offline Object Matching and Evaluation Process for Verification of Autonomous Driving”, ITSC, 2016. 2- J. Martinsson, N. Mohammadiha, A. Schliep, “Clustering Vehicle Motion Trajectories Using Finite Mixtures of Hidden Markov Models”, submitted. 3. E. L. Zec, N. Mohammadiha, A. Schliep, “Modelling Autonomous Driving Sensors Using Hidden Markov Models”, submitted. 4. E. Karlsson, N. Mohammadiha, “A Statistical GPS Error Model for Autonomous Driving”, IV 2018.

Planned approach and implementation

During the recent years, many companies have increased their efforts in developing autonomous driving. One important part of this is related to verification and validation. Machine learning ans statistical signal processing play an important role in this aspect and this project has been able to answer some of the open questions on which models are more suitable for such analysis but it has also shown some new challenges that were not clear from beginning. New projects and activities have to be planned to fully overcome c´such challenges.