Model-based Cartesian Impedance Control (Phase 3)

Anonymous

Keywords: impedance control, Bayesian optimization

TL;DR: Model-based Cartesian Impedance Control for the Real Robot Challenge

Abstract: For this last phase, we adapted our model-based cartesian impedance controller in three ways: An improved pre-grasping position controller, refined grasping points and obtaining new hyperparameters through Bayesian optimization on the real system. Combined, these three adaptations enabled us to tackle the significantly more challenging cuboid object, which required more delicate manipulation compared to the cube. Videos can be found at: https://sites.google.com/view/robotchallenge-modelbased