Title

AWS re:Invent 2023 - Building a simulation platform for AD testing with Torc (AUT206)

Summary

Farooq Khan, Senior Manager at AWS, and Karthik Krishna, Strategy Leader for Autonomous Mobility at AWS, along with Justin Brown and Matt Green from Torc Robotics, presented on building a simulation platform for autonomous driving (AD) testing.
They discussed the challenges in autonomous driving, such as data management, toolchain complexity, and time to market.
AWS infrastructure supports the storage, management, and processing of petabytes of data collected by vehicles.
AWS and partners are working on integrating tools for an end-to-end toolchain to enable seamless collaboration among teams.
The session highlighted the importance of simulation in AD testing due to the impracticality of real-world testing to achieve statistical confidence.
Karthik Krishna explained the historical significance of simulation and its modern applications across various industries.
AWS offers massive compute scale for simulations, with two approaches: scaling up and scaling out.
Justin Brown and Matt Green from Torc Robotics shared their case study on scaling simulations using AWS services.
Torc's virtual driver stack is containerized, allowing deployment flexibility and rapid iteration.
They leveraged AWS services like Lambda, Step Functions, EC2, and others to scale to millions of simulations.
The session concluded with a call to action for attendees to engage with AWS for support in their autonomous mobility development workflows and to provide feedback on the session.

The complexity of autonomous driving increases significantly with each level of autonomy, as defined by the Society of Automotive Engineers, from level 0 (no automation) to level 5 (full automation).
AWS's role in supporting autonomous driving companies is crucial, particularly in managing the vast amounts of data generated and in providing the infrastructure for simulation testing.
Simulation is a critical component in the development and testing of autonomous vehicles due to the need for high confidence in safety without the practical ability to test billions of miles on the road.
The case study from Torc Robotics demonstrates a real-world application of AWS services to solve the challenges of scaling simulations for AD testing.
Torc's approach to containerization and the use of AWS managed services for orchestration and scaling highlights the benefits of cloud infrastructure in handling large-scale, complex computational tasks.
The session underscores the importance of collaboration between AWS and its customers, as well as the continuous need for innovation and improvement in the field of autonomous mobility.