CarND-Controls-PID

Self-Driving Car Engineer Nanodegree Program

PID Controller

A controller takes as input(s) sensor data and convert it/them into an output that is directly controlling an actuator to impact an object’s position or state in the physical world.

In this case, I’m controlling a car in a simulator. Here is a capture of it:

Input:

• Cross track error cte (distance of the car form the center of the road).

Output:

• Steering angle st

A PID controller has 3 components/gains that will be taken into account to calculate the output steering angle:

• proportional Kp
• differential Kd
• integral Ki

The formula is really simple:

st = - Kp * cte
     - Kd * (cte - prev_cte)
     - Ki * sum(cte)

Where prev_cte is the cte at the previous iteration and sum is pseudo code for the sum of all the cteobserved since the beginning of the simulation.

Here is a visualization of the incremental addition of the differential and integral gains.

Parameter tuning

I started with similar values as in the Udacity lessons for the 3 different gains:

Kp = 0.3
Kd = 3
Ki = 0.004

I observed that even though the car would drive continuously, there was too much oscillation. By looking at the first figure, I concluded that the oscillation was due to a proportional gain being to high. Therefore I lowered it to 0.15.

Then, I observed that the correction for an increase in the cross track error would overshoot and result on having the car bouncing on the other side of the road. I deduced that the integral component was the faulty one here. After a few tests, I obtained a value for the integral gain of 0.0008.

I tried tweaking the differential gain without observable improvement.

Final parameters

Kp = 0.15       ///< proportional gain
Kd = 3          ///< differential gain
Ki = 0.0008     ///< integral gain

Dependencies

• cmake >= 3.5
• All OSes: click here for installation instructions
• make >= 4.1(mac, linux), 3.81(Windows)
  • Linux: make is installed by default on most Linux distros
  • Mac: install Xcode command line tools to get make
  • Windows: Click here for installation instructions
• gcc/g++ >= 5.4
  • Linux: gcc / g++ is installed by default on most Linux distros
  • Mac: same deal as make - [install Xcode command line tools]((https://developer.apple.com/xcode/features/)
  • Windows: recommend using MinGW
• uWebSockets
  • Run either ./install-mac.sh or ./install-ubuntu.sh.
  • If you install from source, checkout to commit e94b6e1, i.e.

    git clone https://github.com/uWebSockets/uWebSockets
    cd uWebSockets
    git checkout e94b6e1
    

    Some function signatures have changed in v0.14.x. See this PR for more details.

• Simulator. You can download these from the project intro page in the classroom.

There’s an experimental patch for windows in this PR

Basic Build Instructions

1. Clone this repo.
2. Make a build directory: mkdir build && cd build
3. Compile: cmake .. && make
4. Run it: ./pid.

Tips for setting up your environment can be found here