Understanding Motions of Silicon Robot Through Quality Diversity Algorithms

Simulation is now playing an important role in scientific fields, because we are not able to run physical tests for every assumption and every factor. In this project, instead of attempting to achieve a single task assigned by the scientists-like locomotion in one direction and find the best route, we cared about how diverse the robots behavior can be and how they conduct these behaviors. In order to find the movement capabilities of each robot, the Quality Diversity(QD) Algorithm is applied. By using Multi-dimensional Archive of Phenotypic Elites (MAP-Elites) algorithm as a controller for the material parameters in a simulator, we observed the diversity of behaviors and we believe that it can help us build the repertoire of voxel based robots, which can be used in future to solve assigned tasks on it own.