Alexander L. Burka
aburka (at) seas (dot) upenn (dot) edu
(Facebook)    (Google+)    (Twitter)

CS81 Final Project: Neato Quadcopters

This project pulled together several technologies to try to evolve a neural quadcopter controller. We used ROS and Gazebo, with Team Hector Darmstadt's packages, for a physically realistic simulation of a flying quadcopter. Our software tied this into the NeuroEvolution of Augmenting Topologies (NEAT) algorithm, which evolves the topology and weights of a neural network simultaneously.

Due to likely bugs in our simulation and training code, as well as the combined difficulties of getting neural networks and genetic algorithms to behave, the goal of training a working quadcopter controller was not achieved.


Code and report