Gradient Descent - based motion planning playground

This is an attempt to generalize methods used in an autonomous driver of a virtual racing car described in my Bachelor's thesis.

Installation

Install system dependencies:

• C++14 capable compiler
• cmake
• boost
• sdl2, sdl2-ttf
• fortran77, dh-autoreconf (dependecies of the adept library)

On Debian-based systems you can do it by:

sudo apt-get install libboost-all-dev libsdl2-dev libsdl2-ttf-dev dh-autoreconf fort77 cmake

All other dependencies will be installed within the project build directory by cmake.

Clone repository & build it

git clone https://github.com/acriaer/gokitty.git
mkdir gokitty/build
cd gokitty/build
cmake ..
make #or make -jΦ to build with Φ threads

Usage

./build/demo_app --track=data/tracks/forza.xml --gui=false --port=6001

Try to guess how it works by reading res/default_configuration.xml and the code until I write a comprehensible README.

Software used

• Adept -- used for all the maths, including automatic differentiation.
• SDL2pp -- C++ bindings for SDL2. Used for visulisations.
• CMakeRC -- robust resource compiler for cmake.
• pugixml -- XML library.
• spdlog -- logging (duh).
• boost -- Standard Library for C++.
• TORCS -- racing car game simulator.