Visual servoing of an inverted pendulum on cart using a mounted smartphone

Abstract

Over the last several decades, visual servoing-a vision-based control approach-has been explored as a popular and inexpensive contactless measurement alternative for a variety of industrial robotic systems. Moreover, recent years have witnessed rapid advancements in mobile device applications that have pushed the envelope in the capabilities of smartphone cameras. In this paper, an eye-in-hand pose-based visual servoing (PBVS) approach is presented wherein a smartphone mounted to an inverted pendulum on cart (IPC) system measures both the translational position of a motorized cart and the angular orientation of a pendulum arm for the purpose of feedback control. To perform vision-based control of the IPC system, a discrete-time linear quadratic gaussian (LQG) controller is implemented. Experimental results are presented to characterize the relationships between the frame rate and image resolution of the smartphone camera, processing and wireless communication delays, the measurement noise, and the performance of the closed-loop system.