It's an AGV that runs on a raspberry pie. Recognize the QR coordinates with the lower camera and correct the orientation with the front camera.
The server used the Heroku site, and the app production used Kodular.
-
Replace the pie cam with a wide-angle lens (Due to the actuator space above, the lens is narrow, so the QR coordinate recognition rate is low)
-
DC motor gear ratio greater torque than 1/100 (When AGV lifted the item, I felt that the 1/100 gear ratio I used was not strong enough)
-
When QR coordinates are recognized by Pi-Cam, Pi-Cam calculates the slope of the QR picture and adds an angle correction function (Limited angle correction with Huskylens sensor in front)
-
Highlight bottom ballwheel offset (In the cad file I posted, two ball wheels and two wheels are designed to fit perfectly with the ground, so if the floor is uneven, it doesn't move. So I removed one of the ball wheels in front of AGV)