Summary
In this chapter, we discussed AV and ADAS systems at a high level, along with a reference architecture to build some of these systems on AWS. We also discussed the three main challenges practitioners face when training AV-related ML models in the cloud: feeding TB or more of training data to ML frameworks running on a large-scale, high-performance computing infrastructure, elasticity to linearly scale compute infrastructure to thousands of accelerators leveraging high bandwidth networking, and orchestrating the ML framework training.
Lastly, we walked you through examples of how you can make use of tools on AWS to run SITL simulations for testing your ML models.
In the next chapter, we will focus on solving numerical optimization problems on AWS.
