With AWS DeepRacer, you can create your own machine learning models (in a process called 'training') and race them (in a process called 'evaluation'). You pay for training, evaluating and storing your machine learning models. Charges are based on the amount of time you train and evaluate a new model and the size of the model stored. Additionally, you can purchase a fully autonomous 1/18th scale DeepRacer car, which will allow you to try your models out on real tracks. No purchase is necessary to enter the DeepRacer League.
Announcing the AWS DeepRacer Evo: AWS DeepRacer Evo includes new features and capabilities that help developers to learn more about ML through the addition of sensors that enable object avoidance and head-to-head racing. Starting today, while supplies last, developers can purchase AWS DeepRacer Evo for a limited time discounted price of $399 (38% off MSRP) and AWS DeepRacer Sensor Kit for $149 (43% off MSRP). Both are available on Amazon.com for shipping in the USA only.
To get you started with AWS DeepRacer, you will receive 10 free hours to train or evaluate models and 5GB of free storage during your first month. This is enough to train your first time-trial model, evaluate it, tune it, and then enter it into the AWS DeepRacer League. This offer is valid for 30 days after you have used the service for the first time.
|AWS DeepRacer Service||Price per Unit|
|Training or evaluation||$3.50 per hour|
|Model storage||$0.023 per GB-month|
If you are interested in testing your model’s performance in the real world, visit Amazon.com (US only) and choose between:
AWS DeepRacer ($399) is a fully autonomous 1/18th scale, four-wheel drive car designed to test time-trial models on a physical track. Using a single 4 megapixel camera with 1080p resolution to view the track and a reinforcement learning model to control throttle and steering, the car shows how a time-trial model trained in a simulated environment can be transferred to the real-world.
AWS DeepRacer Evo ($598) is the next generation in autonomous racing. It comes fully equipped with stereo cameras and LiDAR sensor to enable object avoidance and head-to-head racing, giving developers everything they need to take their racing to the next level. In object avoidance races, developers use the sensors to detect and avoid obstacles placed on the track. In head-to-head, developers race against another DeepRacer on the same track and try to avoid it while still turning in the best lap time. Forward facing left and right cameras make up the stereo cameras, which helps the car learn depth information in images. This information can then be used to sense and avoid objects being approached on the track. The LiDAR sensor is backward facing and detects objects behind and beside the car. Developers who already own a DeepRacer can upgrade their cars to have the same capabilities as Evo with the AWS DeepRacer Sensor Kit ($249).
There is no charge for driving your AWS DeepRacer car.
Pricing example #1
When getting started with AWS DeepRacer, developers need to train a reinforcement learning (RL) model. The AWS DeepRacer service will guide you through creating your first model with a series of sensible defaults, followed by training in the virtual simulator. We recommend that you train your time-trial model for at least two hours using the default settings to ensure best chance of convergence, but training time may change if you adjust the parameters.
After training completes, you will use the AWS DeepRacer service to evaluate the performance of your trained model by allowing it to drive autonomously around the virtual track. This allows you to benchmark the performance of your model. The cost of these activities is listed below:
In this example, the total cost is $7.38, assuming you store the model for a whole month. Model evaluation and data stored are estimated because the parameters of your job may impact running time, and may impact data generated. Data stored refers to model checkpoints, metadata, and other files needed to provide the AWS DeepRacer service. You can download the models, to your local drive to save costs, and reimport them at a later stage to continue using them.
Pricing Example #2
While the model you created in Example #1 is a good start, you think you can improve its driving behavior and get a better lap time as a result. You decide to iterate on your model parameters and reward function, and train a new model for 2 hours. The evaluation looks promising but you decide to train it some more, by consecutively cloning it for three times, and training each clone for two hours with small changes in the hyperparameters. The total training time was 8 hours, and you have four models. You evaluate all four versions of the model, keep the best model in your AWS DeepRacer account, and download the three slower models and delete them from your AWS DeepRacer account. You then submit the best model to the AWS DeepRacer League. Racing in the League, or any community race is free.
In this example, the total cost is $29.25.
Pricing Example #3
You decide to leave the models from examples #1 and the best model from example #2 in your AWS DeepRacer account, and come back month over month to submit them to the League. The following table shows the cost for a month of storage assuming you trained your models at the start of a 31-day month.
In this example, the total cost of storage per month is $0.18.
Information on getting started with AWS DeepRacer.
AWS Information on AWS DeepRacer League rules and eligibility requirements.
Get hands-on with RL, experiment, and learn through autonomous driving.