Udacity CarND Path Planning with MATLAB

This is an implementation of the Term 3 project "Path Planning" using MATLAB.

I chose to do this because all of my programming experience is in MATLAB, and I find it much easier to work with personally. I am NOT used to object-oriented program. All of my work is functional programming. I apologize if my poor implementation of classes offends you!

(I also like how I can use logical vectors and matrices to make a bunch of decisions simultaneously without using a for loop every two lines)

Feel free to contact me (tkrebs@ryerson.ca) if you have questions or comments.

Install

• Install MATLAB
• Install jebej/MatlabWebSocket following their instructions
  • I installed the /src/ folder to 'C:/MatlabWebSocket/src/', as indicated in START_SERVER.m
• Run START_SERVER.m to start the server
• Run the Udacity Term 3 simulator and it should connect and control the car
• Modify fcnPATH_PLANNING to make changes to how the car is controlled

About it

• Made with MATLAB 2017a but it should work with previous versions.
  • There may be some issues with scalar addition of matrices in earlier versions?
• jebej/MatlabWebSocket - Used to communicate with the Udacity Term 3 Simulator
• Every time the server recieves a message containing telemetry from the simulator, it runs fcnPATH_PLANNING in fcnPATH_PLANNING.m to generate the next (x,y) points for the simulator. This is where the heart of the project is!
• Right now it is just a basic implementation of the project, using a lot of the methods outlined in Udacity's walkthrough and Q&A and the stuff in my own C++ project (https://github.com/travis-k/CarND)

Good:

• The car stays in its lane usually.
• It shouldn't hit the car in front.
• If there is a car in front, it looks to the left and right to determine lane change safety
• If safe, it will change lanes

Bad:

• The speed adjustment with a car in front is brutally crude, so if a car less than 30 metres in front, then the jerkiness is bad
• Lane changes don't take into account the speed of the car, or the speed of the cars in the other lane
• It only looks to lane change when the boolean too_close is true, which shouldn't be the case necessarily
• Lane changes left or right are chosen at random basically
• The car can get confused if two cars are in front going the same speed, this is the only time I got any warning (out of lane)
• It doesn't return to the center lane when done passing