AWS DeepRacer FAQs

DeepRacer is the fastest way to get rolling with reinforcement learning (RL), literally, with a fully autonomous 1/18th scale race car driven by reinforcement learning and 3D racing simulator. Anyone can can train, evaluate, and tune RL models in the online simulator, deploy models onto an AWS DeepRacer device or RC car for a real-world autonomous experience or compete against each other.

The DeepRacer simulator provides a tutorial to get you started with reinforcement learning and training your first model. You will then also be able to evaluate and tune your models. The DeepRacer Developer Documentation provides additional details on building your first model and also how to improve your models.

No. You can train models, evaluate them without owning an AWS DeepRacer car. 

No. AWS DeepRacer uses the deployed RL model and input from the camera to run inference locally. AWS DeepRacer must be connected to the same Wi-Fi network as the device used to start and stop autonomous driving. Details on how to set up your vehicle can be found in the Developer Documentation.

DeepRacer on AWS integrates with Amazon SageMaker for reinforcement learning model training, AWS RoboMaker to provide the racing simulator, Amazon Kinesis Video Streams for video streaming of virtual simulation footage, Amazon S3 for model storage, and Amazon CloudWatch for log capture.

Currently, developers cannot add additional tracks to the DeepRacer simulator. 

No. Training an RL model requires feedback regarding the outcome of actions taken by the model. This feedback loop exists in the DeepRacer simulator, but not in the real-world.

Currently DeepRacer does not support local training.

Yes. You can use the DeepRacer Distributed Training SageMaker Notebook to create and train RL models and deploy these models to your AWS DeepRacer car manually.