Using evolutionary algorithms to learn a musculoskeletal model balance and walk
The goal of this project is to use evolutionary algorithms for learning musculoskeletal models to walk and control their balance. The EAs develop a neural network controller through generational evolution ending optimizing certain self chosen optimization criteria.
A matlabimplementation of NEAT(Neuro-Evolution of Augmenting Topologies) as evolutionary algorithm. The code is adapted from: http://nn.cs.utexas.edu/?neatmatlab. For details on the NEAT algorithm read the original paper of Kenneth Stanley & Risto Miikulainen: http://nn.cs.utexas.edu/?stanley:ec02