Self-Driving Car Engineer Nanodegree Program

The unscented Kalman filter is a way to improve on the extended Kalman Filter. Unlike the EKF the UKF does not linearize the state equations. It relies on constructing sigma points that get propagated through the state vector model.

Shown below are the results of this project for two datasets.

The noise parameters were chosen in such a way to make the normalized innovation squared close to its statistically expected value. The radar measurement space is three dimensional (rho, phi, rho_dot) and the chi-squared value for a 95% confidence intervall is 7.8. The lidar measurement space is two dimensional (x,y) and the chi-squared value for a 95% confidence intervall is 6. Averaging these two one would expect about 5% of all predicted states to have a chi-squared value of 7 or higher. This is approximately true for the chosen noise parameters.

• cmake >= v3.5
• make >= v4.1
• gcc/g++ >= v5.4

1. Clone this repo.
2. Make a build directory: mkdir build && cd build
3. Compile: cmake .. && make
4. Run it: ./UnscentedKF path/to/input.txt path/to/output.txt. You can find some sample inputs in 'data/'.
   • eg. ./UnscentedKF ../data/sample-laser-radar-measurement-data-1.txt output.txt

Use the following settings:

• indent using spaces
• set tab width to 2 spaces (keeps the matrices in source code aligned)

Please stick to Google's C++ style guide as much as possible.

If you'd like to generate your own radar and lidar data, see the utilities repo for Matlab scripts that can generate additional data.

This information is only accessible by people who are already enrolled in Term 2 of CarND. If you are enrolled, see the project page for instructions and the project rubric.