Gizmo's Life

Proportional Derivative Control

Introduction

Now that we have a robot that can proportionally adjust its wheels based on how close/far it is from the wall, we were asked to add derivative control to our robot.

What it does

The derivative control simply accounts for the difference in error between the current and last reading. Did the last adjustment make much of an improvement in correcting the error? How fast is the error being corrected?

Code

Note: The changes from the previous example having been bolded.

void followWallBeta() {

  delay(500); //wait for us to lift our finger off the button.

  addCSVHeader = true; //for logging data

  leftMotorSpeed=leftMotorDefaultSpeed;

  rightMotorSpeed=rightMotorDefaultSpeed;

  long error;

  long previousError=0;

  while (digitalRead(LineSensor)==HIGH) {

    long pingTime = ping();

    if (pingTime!=0) {

      currentPingDistance = microsecondsToCentimeters(pingTime);

      error = wallDistance-currentPingDistance; //error

      long deriv_error = error - previousError;

      previousError = error;

      rightMotorSpeed=rightMotorDefaultSpeed+(Kp*error)+(Kd*deriv_error);

      leftMotorSpeed=leftMotorDefaultSpeed-(Kp*error)+(Kd*deriv_error);

    } if (pingTime==0) {

      currentPingDistance = microsecondsToCentimeters(pingTime);

    rightMotorSpeed = constrain(rightMotorSpeed, 40, 255);

    leftMotorSpeed = constrain(leftMotorSpeed, 40, 255);

    Forward(0);

    if (useSave) saveData(useSave);

    delay(100); //only check 10 times a second. :)

  Stop();

Results

During our runs, we were having a hard time finding what worked best.

Here the robot followed the wall for a while, until it loses sight of the wall and goes crazy.

VID_20170303_104442.mp4

A video of the above run.

VID_20170303_104602.mp4

Here, the robot continued to go in circles, getting further and further away from the wall with each spin, and never recovered the wall again.

We didn't take any video of this run.

The high spikes are when the robot loses the wall, turns around in a circle, and then reacquires the wall.

This is a closer look at the above trail before it lost sight of the wall

Another day, another wall

Changes made today:

Robots error read is constrained to a maximum of half the target distance
Robot can turn sharper by using negative motor speeds (reversing wheels)

It seems that as we increase the Kd value, the robot performs worse.

Smaller KD Values

The Baseline

No Kd value here. Just some Kp.

At the end the robot still seems to lose the wall.

Google Sites

Report abuse