Stabilization of a Rotary Inverted Pendulum with Neuromorphic Vision Feedback

Abstract

This work aims to stabilize the Furuta pendulum using neuromorphic vision feedback from an event-based camera. A mathematical model is derived from the mechanical design of the Furuta pendulum. A swing-up trajectory is obtained by solving an optimal control problem. The system states are estimated using an Extended Kalman Filter to stabilize the Furuta pendulum about the reference trajectories with an LQR controller. An extension of the traditional Hough transform algorithm with exponential decay for event-based cameras is introduced to obtain the pendulum angle directly from the event stream. The proposed method is validated through experiments on the Furuta pendulum. It is shown that tracking and stabilization of the Furuta pendulum with event camera feedback about reference trajectories is achieved.