For my master thesis I am building a quadruped robot, which basically looks like a robotic cat. The aim of the thesis is to train the robot by using a machine learning technique within evolutionary algorithms, called genetic algorithm.
Each leg of the robot is powered by two servomotors, where the servomotors cyclic position is controlled by coupled central pattern generators (CPG’s). These CPG’s are inspired by how the spinal cord of vertebrates handles locomotion, balance, and reflexes.
The mechanical design of the robot is inspired by work done at the École Polytechnique Fédérale de Lausanne research institute in Switzerland. They analyze animal movement in order to replicate steady and robust robot locomotion.
I will try to optimize the parameters of the CPG’s with machine learning techniques, where the fitness is measured by velocity, stability and power consumption.
The body parts of the robot are 3D printed, which gives a lightweight robot that is easily customizable.