Charingchen/CarND-PID-control

A PID tuner using Twiddle logic/ Gradient descent

CarND-Controls-PID

Self-Driving Car Engineer Nanodegree Program

---

PID Tuning

From Wiki

Parameter	Rise time	Overshoot	Settling time	Steady-state error	Stability
K_p	Decrease	Increase	Small change	Decrease	Degrade
K_i	Decrease	Increase	Increase	Eliminate	Degrade
K_d	Minor change	Decrease	Decrease	No effect in theory	Improve if K_d small

Twiddle logic

I am using a modified twiddle logic in this PID tuner. To achieve run twiddle only once every time gets the error readings, a state machine is used inside of for loop. There are three states, INIT: Initialize best err and update PID values by adding dp values, INCREMENT: Increment dp if a better error found, vis versa, DECREMENT: if fail second times in a row, decrement dp with 0.9

However, I was first thinking to run this logic every time I got a cte from the simulator. This fails because twiddle and PID controller would not react fast enough and result in full left and right turns every step like following showed:

As the result, more errors need to be added together to feed into the twiddle. At beginning of the tuning, I used the average of the error to find a range of PID values and reach the values faster. Later when I have a value, I will use the sum of all errors to feed into the twiddle.

Tuning Process

Starting from P=0,I=0,D=0 and stepping dp 1,1,1. After about 20 iteration of twiddle, I was able get PID value of P:2.5 I:0 D:10.

I was able to observe a lot of overshoot and oscillation at a slight left turn. Like below

Tuning down P and I to 0.5 and 0.0001 gives a better result. D is at 5. However, after this I observed during a sharp turn, the car would oscillate. I decide to increase D to decrease overshoot and increase P a bit to reduce rise time. Then I run twiddle with the initial value of P= 0.531 I = 0.0001 D =9.6 At this point I value should be good, I only run P and D with dp value of [1,1]

After 7 Iteration I get following value

Kp	Ki	Kd
0.8389	1.9e-5	10.6961

Conclusion

In summary, the two values below would all work. I feel the twiddle would prefer more aggressive P and D to achieve faster error corrections. The lower P value should be more pleasant to sit in.

Kp	Ki	Kd
0.8389	1.9e-5	10.6961
0.5	0.0001	9.6

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.

Fellow students have put together a guide to Windows set-up for the project here if the environment you have set up for the Sensor Fusion projects does not work for this project. There's also 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