Computer Vision and AI-based Systems for AVs and the Exponential Data Growth

SAE Innovations in Mobility presents...

Our speakers:

Ralf Nikolaus (Altran)
Lawrence Vivolo (Dell EMC)
Topic: Meeting the Challenges of Exponential Data Growth Head-on

About the Speakers:
Ralf has more than 10 years business experience within Advanced Driver Assistance Systems (ADAS) & Autonomous Driving (AD). He is supporting different automotive clients with a unique service-center approach. He manages a transnational team and coordinates world-wide activities in regards to Altran VueForge® ADAS/AD, paving the way to reliable and safe autonomous driving. He has a Computer Science degree with a focus on Machine Learning and Image Processing.
Lawrence brings more than 30 years in high-technology design, marketing and business expertise to his role as Automotive and Electronic Design Automation (EDA) Business Development Manager for Dell EMC. He holds a BSEE from Cal Poly, San Luis Obispo, and an MBA from Santa Clara University.

About the Presentation:
Depending on the test vehicle and test day, 30-194 terabytes (TB) of data per vehicle and day can be captured. When validating a new generation of vehicle alone, volumes of test data between 15 and 100 petabytes (PB) are collected for SAE level 3. This data needs to be measured with chronological synchronism, saved, organized, processed, analyzed and evaluated. For SAE level 5, the data requirements can grow into multiples of exabytes (EB).
During their presentation, Ralf and Lawrence will describe some of the challenges and best practices that are related to the complexity and volumes of data to efficiently validating driver assistant systems:
• Automated collection of all meta-information and pre-labeling and data enrichment
• Validation and testing of software modules relevant to safety
• Inter-divisional development in non-homogeneous sensor tool landscapes
• Operation of a scalable SiL and HiL environment as well as the execution of SW tests, simulation, analysis and evaluation
• Distributed data management of very large quantities of data (mass data)
• Distributed and parallelized data processing and flexible access to collected and processed data (big data, private cloud computing)
• Data retention and security
• Potential risk of public- vs. private cloud implementations
• Saleable IT infrastructures that can grow dynamically

Altran will share best practices of how the entire test and validation process can leverage a Dell EMC private cloud environment and/or managed service infrastructure solution to address these challenges and present the joint Altran-Dell EMC reference architecture.

Peter Gronerth (FKA)
Topic: Leveraging real world and synthetic data for combined datasets

About the Speaker:
Peter is currently a researcher & developer at fka‘s HQ in Aachen, Germany. He will be joining FKA SV starting January 2019 with the focus of implementing advancements made at the HQ into both the immersive driving simulator and FKA SV’s test vehicle. Originally, Peter started working for the institute for automotive engineering (ika) at RWTH Aachen University in 2009 as student researcher and joined full time after earning his M.Sc. in Electrical Engineering in 2015. His research focuses on digital image processing and machine learning, both essential to his work regarding the creation of environment models for autonomous driving.

About the Presentation:
During this presentation Peter will give insights into current developments made at IKA and FKA as well as into his own research. The later one concentrates on the creation of
semi-artifical training data applied to train neural networks. The resulting datasets leverage the strengths of real world data (available e.g. through KITTI, CamVid, Berkley DeepDrive or Cityscapes)
and artifical datasets (e.g. Synthia) by recombining elements of available images.

Starting at 6:30pm, food and drinks as always. There will be time for Q&A and a chance to try fka SV's driving simulator.

See you there,
John & Christian