AWS DeepRacer is a 1/18th scale race car which gives you an interesting and fun way to get started with reinforcement learning (RL). RL is an advanced machine learning (ML) technique which takes a very different approach to training models than other machine learning methods. Its super power is that it learns very complex behaviors without requiring any labeled training data, and can make short term decisions while optimizing for a longer term goal.
With AWS DeepRacer, you now have a way to get hands-on with RL, experiment, and learn through autonomous driving. You can get started with the virtual car and tracks in the cloud-based 3D racing simulator, and for a real-world experience, you can deploy your trained models onto AWS DeepRacer and race your friends, or take part in the global AWS DeepRacer League.
The platform consists of three parts:
AWS DeepRacer is an autonomous 1/18th scale race car designed to test RL models by racing on a physical track. Using a camera to view the track and a reinforcement model to control throttle and steering, the car shows how a model trained in a simulated environment can be transferred to the real-world. The combination of console, simulator, and car provides a complete solution to experiment with RL algorithms and generalization methods.
Build models in Amazon SageMaker and train, test, and iterate quickly and easily on the track in the AWS DeepRacer 3D racing simulator. With AWS DeepRacer you do not need to manually set up a software environment, simulator or configure a training environment. AWS DeepRacer offers an integrated simulation environment and reinforcement learning platform hosted on the AWS Cloud for experimentation and optimization of your autonomous racing models.
Compete in the world’s first global, autonomous racing league, to race for prizes and glory and a chance to advance to the Championship Cup. Every month, there are races on the simulator, combined with in-person events spread throughout the year.
Our organization, ARCC, has studied alternatives for learning and teaching the AWS platform.
- Integrated Approach
- Simulation Based - No expensive car or track required
- Abstracts most of the software: Allows the developer to focus just on the RL model
- Competition - Unifies community into large scale competition
- Active development community
- Standardized car, race is just about software
- Uses ROS on backend
- AWS platform can get expensive
- Not much flexibility in ML models, you're stuck with PPO
There are a few other alternatives to the AWS DeepRacer if you're looking to race autonomous cars
The Donkey Car is a community project to opensource self driving cars. They also offer kits to build the car seperately. All in all, the car cost around $250, which prices it lower than the DeepRacer. Theres also no standardization in cars, every car has a different chasis and motors, meaning hardware matters just as much as software. Additionally, DonkeyCar isn't supported by an organization like Amazon, so it's a little more spread out and not as organized compared to DeepRacer.
- Actually cheaper
- Completely Open Sourced
- Pretty Large Community
- Community isn't as big as AWS DeepRacer
- Competitions are spread out
- No simulator to test and develop in, requires large track
The group I work with, ARCC, built a self driving car from scratch using a Traxaas RC car, NVIDIA Jetson TX2, and Intel Realsense camera. This was our first project in autonomous cars, and we still develop on it with alternative algorithms and methods.
- Fast
- Much more powerful hardware
- Fully customizable - Choose whatever hardware or software stack you want
- Hardware problems to debug :(
- Much more expensive
- No community support as it's a oneoff
- Too big for most tracks, we race it on a 400m running track