CycleIK: Neuro-inspired Inverse Kinematics

Abstract The paper introduces CycleIK, a neuro-robotic approach that wraps two novel neuro-inspired methods for the inverse kinematics (IK) task—a Generative Adversarial Network (GAN), and Multi-Layer Perceptron architecture. These can be used in standalone fashion, but we also show how embedding these into hybrid neuro-genetic IK pipeline allows further optimization via sequential least-squares programming (SLSQP) or genetic algorithm (GA). models are trained tested on dense datasets were collected from random robot configurations of new Neuro-Inspired COLlaborator (NICOL), semi-humanoid with redundant 8-DoF manipulators. We utilize weighted multi-objective function state-of-the-art BioIK method to support training process our neural compete approaches, which deployment directly robotic hardware. Additionally, it is shown incorporation improves precision while simultaneously reducing overall runtime.