A Brain Inspired Learning Algorithm for the Perception of a Quadrotor in\n Wind

Abstract

The quest for a brain-inspired learning algorithm for robots has culminated\nin the free energy principle from neuroscience that models the brain's\nperception and action as an optimization over its free energy objectives. Based\non this idea, we propose an estimation algorithm for accurate output prediction\nof a quadrotor flying under unmodelled wind conditions. The key idea behind\nthis work is the handling of unmodelled wind dynamics and the model's\nnon-linearity errors as coloured noise in the system, and leveraging it for\naccurate output predictions. This paper provides the first experimental\nvalidation for the usefulness of generalized coordinates for robot perception\nusing Dynamic Expectation Maximization (DEM). Through real flight experiments,\nwe show that the estimator outperforms classical estimators with the least\nerror in output predictions. Based on the experimental results, we extend the\nDEM algorithm for model order selection for complete black box identification.\nWith this paper, we provide the first experimental validation of DEM applied to\nrobot learning.\n