Amazon Deepracer is the fastest way to start using Machine Learning (ML). You can use a 1/18 scale Enhanced Learning (RL) model for self-driving vehicle training in a cloud-based virtual simulator and compete for prizes and honors in the Amazon Deepracer League worldwide. Today, we will extend the Amazon Deepracer functionality by opening source the Amazon Deepracer Device software to provide fun hands-on learning.
Why open source
Amazon Deepracer’s virtual and offline races have been well received, but developers now want its car to go beyond racing leagues. Amazon Deepracer is an Ubuntu-based computer-wheeled car powered by our open-source robotic operating system (ROS) that allows developers with basic Linux coding skills to easily prototype interesting new uses for their cars. The Amazon Deepracer device software is now publicly available, so anyone with a car and an idea can make a new use of their device a reality.
We’ve compiled six sample projects from the Amazon Deepracer team and members of the global Amazon Deepracer community to help you start exploring the limitless possibilities that open source can achieve. When developers share new projects using # DeepRacer project, we will highlight our favorites on the Amazon DeepRacer robot project page. Whether it’s using DeepBlaster to load a Nerf cannon on your car, creating a virtualized effect for your home or office with Mapping, or proposing new ways to race cars with friends and colleagues with DeepDriver, You can do all of this using open source and sample projects. The documentation is available on GitHub and can be collaborated with thousands of community members on the Amazon Deepracer Slack channel. The only limit to the Amazon Deepracer’s potential is your imagination (and, of course, the laws of physics).
Let’s start the experiment
With the Amazon Deepracer device code open source, you can quickly and easily change the default behavior of a car on the track you are currently tracking. Want to prevent other vehicles from overtaking by deploying countermeasures? Want to deploy your own custom algorithm to increase the speed at which A car gets from point A to point B? You just have to think, and then you can code. We’d love to see your ideas, from new forms of racing cars to new uses for Amazon Deepracer.
Starting today, you can choose from six projects (Follow the Leader, Mapping, Off-Road (created by Amazon Web Services), RoboCat, DeepBlaster, and DeepDriver (created by the open source community), or create a project of your own. You can start with the Follow the Leader example project, which will train the car to detect and track an object. This is the fastest build and run project, and in the next section, we’ll show how easy it is to modify the default behavior of an Amazon Deepracer car. To complete this setup, upgrade to the latest software version and get into your car via SSH.
Updated version link:
https://docs.aws.amazon.com/d…
Download the Follow the Leader project
Connect to the car using SSH, switch to root, and create a working directory. Then clone the “Follow the Leader” GitHub repository:
sudo su
mkdir -p ~/deepracer_ws
cd ~/deepracer_ws
git clone
https://github.com/aws-deepra…
The process of fully cloning the project repository into the car may take a few minutes (depending on your network connection speed). The Follow the Leader project includes several installation scripts to help simplify the process of getting up and running faster. In addition, if you are more accustomed to running shell-based commands or want to see the process of using the relevant documentation for each stage in more detail, you can also do the following steps manually.
Download and transform the object detection model
First, we need to download and transform the object detection model. To do this, we can run the script that comes with the Follow the Leader repository:
sudo su
cd ~/deepracer_ws/aws-deepracer-follow-the-leader-sample-project/installers
/usr/bin/bash install_object_detection_model.sh
The installer script will download and tune the model, and then copy the tuning project to the model location. This process should take about 3-4 minutes.
You can do this phase manually using the detailed instructions for downloading and transforming the object detection model:
https://github.com/aws-deepra…
If it has not been initialized previously, then rosdep is initialized
Rosdep helps install dependency packages. If it has not been initialized on the device before, then ROSDep is initialized first.
sudo rosdep init
sudo rosdep update
Build the Follow the Leader package
Next, we need to extract the package dependencies required by the project and build them:
sudo su
cd ~/deepracer_ws/aws-deepracer-follow-the-leader-sample-project/installers
/usr/bin/bash build_and_install_ftl_application.sh
After success, you should be able to view a screen similar to the following:
The script downloads and installs the required package dependencies, and builds the package. This process may take about 8-10 minutes to complete.
You can also complete this phase manually by following steps 1-10 in the “Download and build” section of Leader README. The installation script will do the same (just save you some typing).
Launch the Follow the Leader application
Now we can run the Follow the Leader application:
sudo su
cd ~/deepracer_ws/aws-deepracer-follow-the-leader-sample-project/installers
/usr/bin/bash run_ftl_application.sh
Enable “Follow the Leader” mode
Finally, we need to open another SSH session for the vehicle using the command line interface (CLI) to enable the “Follow the Leader” mode:
sudo su
cd ~/deepracer_ws/aws-deepracer-follow-the-leader-sample-project/installers
/usr/bin/bash enable_ftl_mode.sh
Now, either you or the volunteer (or an object) can move and watch the traffic begin to follow! Isn’t that great?
Share your results
A: congratulations! You have completed your first sample project. Share your experience with friends and family on social media using the hashtag # DeepracerProject to let us know how your work is going. As the community creates more Amazon Deepracer projects, we’ll add them to the Amazon Deepracer GitHub organization and detail them in a subsequent blog post so everyone can get inspired.
The resources
Amazon DeepRacer:
https://aws.amazon.com/deepra…
Amazon Deepracer Robot Project:
https://aws.amazon.com/deepra…
DeepBlaster:
The Mapping:
DeepDriver:
Making:
https://github.com/aws-deepra…
Amazon Deepracer Slack channel:
https://deepracing.io/
Follow the leader:
Cross-country:
https://github.com/aws-deepra…
RoboCat:
Connect to the car using SSH:
https://docs.aws.amazon.com/d…
Follow the leader README.md:
https://github.com/aws-deepra…
David Smith, Senior Solution Architect at Amazon Deepracer
He is passionate about Amazon Deepracer, technical support, and learning. Outside of work, he enjoys Formula 1, drone flying (and crashing), 3D printing, running (running in the park), tinkering with code and spending time with his family.
[Amazon Cloud Technology China Summit in 2021] Take you to feel the four-wheel drive brother era! This summit will be grandly held in Shanghai, Beijing and Shenzhen. Together with many leading technology practitioners in the industry, they will share multi-dimensional forward-looking perspectives around the reliable technology and service of Amazon Cloud Technology. The “Amazon Deepracer Racing Circuit” was specially created in the exhibition area, so that you can get close to the AI front line and the frontier of machine learning development.
More information about the summit can be found herehttps://aws.amazon.com/cn/abo…
What are you waiting for? Sign up for a marathon with Amazon Deepracer in a crazy self-driving race! Let’s salute the four-wheel drive boy with AI!