Servo Controller

This controller runs on a Raspberry Pi Pico microcontroller. The Pico may optionally be soldered to a main PCB that facilitates interconnects with up to 16 RC servos, 2 power supplies and a computer.

This controller is designed to use the "Precision Servo System" that runs on a Linux computer. This system includes a GUI app for servo calibration. This app allows for the robot's as-built joint angles to be measured and does a least squares fit of these measurements to find a function that transforms joint angles in radians in the joint's frame to microseconds of pulse width for the servo that drives that joint.
This calibration corrects for differences between each servomotor, and inaccuracies in the installation of the servo. The users can also set soft limits for joint angles. This data is stored in a calibrations file.

In addition to setting the position the user can set a rate in radians per second. The servo's position will be updated so as to make it move at the specified rate.
It will continue to move untill it either hits a soft limit or the rate and postion is updated.
The default rate is zero.

This servo controller can store the data from the calibration file and apply it in real time. So the Linux computer sends the desired angles in radians in the joints frame to the Pico that in turns generates the PWM pulses for up to 16 servos.

The Precision Servo System can also use a PCA9685 breakout board to drive a set of up to 16 RC servos. This controller replaces the PCA9685 board and does several things better:

1. The PWM pulses are much more accurate.
2. An angular rate can be specified so the robot joint can move independently of position updates from the Linux computer
3. The PCA9685 uses I2C while the Pico has a few other options to connect with the Linux PC

Future Work

I may move this software to a Pico-W or ESP32. Both of these have WiFi radios that might allow the servos to be controlled over a wireless link. This would allow the Linux computer to be removed from the robot, allowing for a smaller and lower priced robot. (Implementing the above angular rate feature should reduce WiFi bandwidth and importantly allow for smooth motion if WiFi packets are dropped.)

I may implement inverse kinematics in this controller so that the Linux PC only has to send the desired (x, y, z) locations and optional velocities of the end points.
My guess is that the ESP32 would be used because it has much faster floating point hardware.

Author

Chris Albertson [email protected]