russosk1

Looks pretty similar to other builds I saw)
 Img1.png   3,15 Mo   0 téléchargement(s)
 Img2.png   3,36 Mo   0 téléchargement(s)

Hi everyone! I am finally sharing what is achieved so far!

The robot is built around **two computers**, each handling the tasks they are best suited for:

**Raspberry Pi 5**

* Human detection and recognition.
* Controls the motors and overall robot movement.
* Acts as the main controller for navigation.

**RDK X5**

* Computer vision processing.
* Depth perception using a stereo camera.
* Object detection and environmental scanning.

I decided to split the workload this way because it keeps each board focused on what it does best. The Raspberry Pi handles the robot's movement and interaction, while the RDK X5 performs the more demanding vision tasks and sends the results back to the Raspberry Pi. The RDK X5 runs ROS 2 for the vision pipeline, handling stereo depth estimation and object detection. The processed results are transmitted over Ethernet to the Raspberry Pi, which listens for these updates and translates them into motor commands and navigation decisions. This separation allows the RDK X5 to focus entirely on AI inference while the Raspberry Pi remains dedicated to real-time robot control.

At this point, the communication between both boards is working, human recognition is working, motor control is working, and the RDK X5 is successfully processing stereo vision and object detection. It's really satisfying to see everything finally working together!

Right now I'm working on making the robot completely autonomous by adding a battery pack and an onboard charging system, so it no longer needs to stay connected to the wall.

The next big milestone is autonomous charging. The plan is for the RDK X5 to detect the charging station with a QR, use its depth perception to estimate the station's position in 3D space, and guide the Raspberry Pi so the robot can drive itself to the charger and dock automatically when the battery is low.

I'm looking forward to sharing more progress soon, and I'd love to hear any suggestions or ideas from the community!