Build a custom AWS DeepRacer car
You can build a custom vehicle that is fully compatible with the DeepRacer on AWS solution. You can follow three build paths supported by the AWS DeepRacer Community open-source project
-
Option A—Replace the Intel Atom processor in an existing AWS DeepRacer Evo with a Raspberry Pi 5, servo driver HAT, and Raspberry Pi-compatible camera, while reusing the existing chassis and wheel/motor assembly.
-
Option B—Build a new vehicle from scratch using a 1/18-scale RC truck chassis, Raspberry Pi 5, servo driver HAT, camera, and 3D-printed mounting components.
-
Option C—Install a custom operating system on an original AWS DeepRacer device (Intel Atom-based) using the developer bootloader shim, without any hardware modifications.
Options A and B use the same software installation procedure. Option C uses the developer bootloader shim to install a modern operating system. Custom cars that you build with any option are fully compatible with reinforcement learning models that you train using the DeepRacer on AWS solution.
The following table compares the three build options to help you decide which path is right for you.
| Category | Option A: Upgrade existing DeepRacer Evo | Option B: Build from scratch | Option C: Install a custom OS on original device |
|---|---|---|---|
|
Best for |
Customers who have an existing AWS DeepRacer Evo car. This is the fastest and easiest path. |
Customers who do not have an existing AWS DeepRacer car and cannot procure one. This is the only available path in that case. |
Customers who have an original AWS DeepRacer device (Intel Atom-based) and want to install a modern operating system using the developer bootloader shim. |
|
Compute module |
Raspberry Pi 5, servo driver HAT, and camera |
Raspberry Pi 5, servo driver HAT, and camera |
Original Intel Atom compute module (existing device hardware) |
|
Software |
Same installation procedure for both options |
Same installation procedure for both options |
Developer bootloader shim with custom OS installation (Ubuntu 24.04 community distribution, third-party OS, or custom OS) |
|
Model compatibility |
Fully compatible with reinforcement learning models trained using the DeepRacer on AWS solution |
Fully compatible with reinforcement learning models trained using the DeepRacer on AWS solution |
Fully compatible with reinforcement learning models trained using the DeepRacer on AWS solution (when using the community distribution) |
|
Batteries |
Single battery |
Two batteries |
Existing device battery |
|
Chassis |
Reuses existing chassis, ESC, servo, and wheel/motor assembly |
Requires a 1/18-scale RC truck chassis and off-the-shelf components |
Uses existing original AWS DeepRacer chassis |
|
Soldering |
Not required |
Required (soldering iron needed) |
Not required |
|
ESC knowledge |
Not required |
Required (you must swap the Electronic Speed Controller) |
Not required |
|
3D printing |
Required — front camera mount, main plate (with legs), rear mount, and body mounts (×2) |
Required — front camera mount, main plate (without legs), rear mount, body mounts (×2), and additional chassis-specific mounting components |
Not required |
|
Approximate cost |
$220 |
$350 |
$0 (uses existing device) |
|
Estimated assembly time (excluding 3D printing) |
2 hours |
3–5 hours |
30 minutes |
