- Notifications
You must be signed in to change notification settings - Fork3
A Nvidia partnership project of a autonomous car MVP using Jetbot and other Nvidia tools for HPC and Transfer Learning
pfeinsper/NVIDIA
Folders and files
| Name | Name | Last commit message | Last commit date | |
|---|---|---|---|---|
Repository files navigation
This is an open-source project of a MVP (Minimum Valiable Product) of a autonomous car usingJetson Nano that respects brazilian traffic signs and integrates some Nvidia tools, such asDeepStream SDK,Nvidia GPU Cloud,TensorRT,CUDA Toolkit,Triton andJetbot as the main tool for the project.
The main workflow of the project shown on this figure above. The architecture of the project is based on four sections:
- Autopilot
- Neural Networks
- Detector
- Broker MQTT
- Controller
For more details on how to start with our project, check our documentation:
So the robot starts following the road using computer vision and the controller parameters to stay on track, as the Jetbot sees any traffic sign, for instance a stop sign, the robot immediately detect the stop sign as a roadsign from the detector model and uses it to detect the roadsign as a stop sign by the traffic sign model. After this, the detector connects with a broker that uses asyncronous message to communicate with the autopilot and tells the robot to stop, so the motors receive a speed gain of zero and the robot stops on the track, respecting the traffic sign identified.
