ADAS Simulation on Cloud

2024-28-0166

To be published on 12/05/2024

Event
11th SAEINDIA International Mobility Conference (SIIMC 2024)
Authors Abstract
Content
Advancements in new technologies in the market, meeting the on-demand requirements of automotive customers to provide best-in-class simulation solutions, help in the faster development of software. Developers face numerous challenges, especially in processing vast amounts of data. For instance, a single test vehicle for an Advanced Driving Assistance System generates nearly thousands of TB of data daily. Managing these data sets and ensuring their timely availability across distributed sites can lead to significant delays in development cycles. Additionally, the high volume of test cases in simulation and validation further contributes to these delays. On-premises simulation can arise several potential problems and limitations, especially for organizations that rely on setting up HPC (High-Performance Computers). Here are some key issues that can be faced: limited computational resources, scalability issues, portability, high capital and maintenance costs, resource management and utilization, compatibility issues between GPU drivers and servers impacting both performance and cost, and security and compliance challenges. Towards adapting cloud-based simulation using AWS services such as S3, Batch, Auto Scaling Group (ASG), and Elastic Container Registry (ECR), simulations are conducted in parallel to test Software-In-Loop (SIL). Multiple compute instances that can run simulations can be launched simultaneously, significantly speeding up the testing process and GPU-based instances can be developed in low-cost instances, which can save 3x the simulation cost. The Docker container built for simulation includes the software algorithm, simulation tool, and necessary recordings. Specific driving functions such as Blind Spot Detection (BSD), Lane Change Assist (LCA), and Occupant Safety Exit (OSE) can be tested. Recordings from actual driving scenarios are used to validate the performance of these functions. Workbench has an innovative environment that enables developers to create software solutions based on virtual ECUs before the hardware is even available. The automotive operating system used for SW development provides strong middleware for a standardized, secured connection from computer to cloud.
Meta TagsDetails
Citation
Ramapuram, V., Dhar, J., and Munaiahgari, M., "ADAS Simulation on Cloud," SAE Technical Paper 2024-28-0166, 2024, .
Additional Details
Publisher
Published
To be published on Dec 5, 2024
Product Code
2024-28-0166
Content Type
Technical Paper
Language
English