Webots simulation of e-puck robot that co-evolves games and solutions.
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.idea
controllers
worlds
.gitignore
README.md

README.md

darwinian-epuck

![Alt text](http://i.imgur.com/0NBekhn.png Robot following line)

Simulation of a Playful Robot with a Darwinian Neurodynamic Brain

October 2012 – April 2013

This project was part of a study on cognitive architecture that could be implemented in a robot. The aim was to implement algorithms that are based on Darwinian Neurodynamics, to perform experiments and identify future improvements. One of the alorithms designs that was implemented was an evolution of Solutions (Artificial Feedforward Neural Network) that learn how to play multiple games. Experiments on all algorithms were performed using Webots robot simulator. The algorithm in this repository was implemented on a simulation of a robot with two differential wheels, E-puck. Experiments on E-puck using a single Game resulted in the targeted behaviour. Implementation of multiple Games resulted in mediocre Solutions and led to modification of the algorithm to use Multi-Objective Optimisation. Further improvements that have been identified include modification of Games to increase their complexity and make them evolvable.

Some videos:

Obstacle avoidance: https://www.youtube.com/watch?v=21bRxwns5eg

Following the line: https://www.youtube.com/watch?v=6ioIH3lnZHE