For the wheeled structure
I built the mechanical structure and developed the software. The robot can now follow a person using a camera. The OAK camera is connected to a Raspberry Pi 4, where a script extracts the 3D coordinates of the detected person’s face and sends them to the STM32, which controls the wheel motors.
The structure is made of metal and features a suspension system that ensures all wheels remain in contact with the ground. It is sufficiently heavy and robust to support the upper structure without being damaged or overturned.
For the hands
As for the arms
Thanks to multiple servomotors, the arm can reproduce all the movements of a real human arm. The arm includes a hollow section that contains a custom-designed PCB, which controls and supplies power to all the servomotors in the hand and part of the arm.
For the waist
The goal of this section is to create a system that can move along three axes, tilt forward and backward, and lean to the right and left.
I could use DC motors or powerful stepper motors, along with gears and springs, to solve this problem. However, due to budget and time constraints, I chose a different approach.
I decided to use a system based on three linear actuators and a triangular structure. With three actuators, it is possible to simulate movement along the three axes without any issues, and it can also simulate walking motion.
I selected actuators with a speed of 40 mm/s and a force of 200 N. A speed of 40 mm/s is sufficient because the three actuators move simultaneously. To double the effective speed, one actuator can move downward while the other two move upward. If I chose faster actuators, they would not be able to support the required weight.
