Lesson: Sensory-Motor Control


Today's Lesson




Sensory-Motor Control and PID Control


One foundational concept in robotics is sensory-motor loops and control. The robot's sensor readings inform how it moves in its environment, which will then influence the next sensor readings from the robot, and the loop goes on. One of the most well known sensory-motor control methods is PID (Proportional Integral Derivative) control, which is depicted in the block diagram below:

PID control

A block diagram of a PID controller in a feedback loop. Source: Wikipedia.

The PID control function can be represented as follows:

PID control equation

Where Kp, Ki, and Kd, are the constant coefficients for the proportional, integral, and derivative terms; e(t) represents the difference between the goal (setpoint) and the sensor measurement (process variable). The three following graphs display how a system responds to a step change in the setpoint to each component of the PID controller separately and all the components combined at different values of Kp, Ki, and Kd.

Proportional control

Response of the process variable to a step change of the setpoint for different values of Kp. Source: Wikipedia.

Integral control

Response of the process variable to a step change of the setpoint for different values of Ki. Source: Wikipedia.

Derivative control

Response of the process variable to a step change of the setpoint for different values of Kd. Source: Wikipedia.

This YouTube video by Brian Douglas is a great resource providing a clear explanation of the PID controller.


Coding Exercise: Line Following


Note: Here's links to files
setup.py
gazebo_lineFollow.launch.py
camera_lidar_gazebo.urdf
line_world.world
launch_config.rviz

Getting Started


Considering, we have created iml_buche package :
$ ros2 launch iml_buche gazebo_lineFollow.launch.py 

In Gazebo, you should see a vehicule at the beginning of a yellow line in an enclosed room (see image below). Your goal is to program the robot to follow the yellow line, writing your code in line_follower.py.

Yellow line Gazebo sim

The starter code implements a helpful debugging window to help visualize the center of the computed yellow pixels. Once you've correctly identified the center of the yellow pixes, your window should look something like the following:

Red dot visualization

Understanding the Starter Code


Read through the starter code in line_follower.py. We encourage you to look up what certain OpenCV functions do to better understand what's going on. Make sure that you all discuss and understand what the following variables represent and what values they will hold: h, w, d, cx, cy.


Implementing the Line Follower


This programming exercise contains 2 main components:

  1. Defining the range for what you will consider a "yellow" pixel in the image feed.
    • OpenCV uses the following ranges for H, S, and V: H: 0-179, S: 0-255, V: 0-255. As this OpenCV documentation suggests, you may find the following helpful (except you'll want to investigate yellow instead of green):
      green = np.uint8([[[0,255,0 ]]])
      hsv_green = cv2.cvtColor(green,cv2.COLOR_BGR2HSV)
      print(hsv_green)
    • You may also find the "Swatches" section on the HSV Wikipedia page helpful. Remember, you'll have to convert from the HSV ranges in the Wikipedia page H: 0-360, S: 0.0-1.0, V: 0.0-1.0 to the numeric ranges OpenCV expects H: 0-179, S: 0-255, V: 0-255.
      • For example, the HSV Wikipedia color \(H = 60^{\circ}, S = \frac{1}{4}, V = \frac{5}{8}\) converts to the OpenCV HSV values of \(H = \big(\frac{60^{\circ}}{360^{\circ}} \cdot 180 - 1\big) = 29, S = \big(\frac{1}{4} \cdot 256 - 1\big) = 63, V = \big(\frac{5}{8} \cdot 256 - 1\big) = 159\)
  2. Implementing proportional control to enable the robot to follow the yellow line. This will involve:
    • Setting up a ROS2 publisher to control the movement of the robot.
    • Computing an "error" term (the difference between the goal and the sensed reality). This most important part here is defining the "goal" and the "reality" comparison.
    • Using the error term to determine how the robot moves in accordance with the principles of proportional control.

To run your code:

$ ros2 run iml_buche line_follower_mode

Once you've successfully implemented your proportional control line follower, it should look something like the following:

line follower demo