Methods to Run Thousands and thousands of Self-Driving Automotive Simulations on GCP (Cloud Subsequent '19)



Cruise Automation is a number one developer of autonomous automobile know-how. On this session, we are going to dive into the infrastructure which permits us to run lots of of 1000’s of autonomous simulations daily and analyze the outcomes rapidly and effectively. Cruise runs the overwhelming majority of our testing on Google Cloud, profiting from excessive scalability of compute and GPU sources for our numerous workloads.

Our simulation frameworks permit us to replay knowledge gathered from highway testing or generate complicated variations on human-defined eventualities. We are going to check out the know-how and infrastructure behind these frameworks, challenges scheduling and scaling our workload on GCP, complexities accessing high-definition 3D maps and different sources required to run every simulation, and our knowledge pipeline which permits engineers to rapidly instrument and analyze their outcomes.

Construct with Google Cloud →

Watch extra:
Subsequent ’19 Compute Classes right here →
Subsequent ‘19 All Classes playlist →

Subscribe to the GCP Channel →

Speaker(s): Adrian Macneil

Session ID: CMP207
product:Compute Engine;

Leave a Reply

Your email address will not be published. Required fields are marked *