State estimation for a holonomic omniwheel robot using a particle filter

Abstract

The holonomic robot platform designed for the Opera of the Future must perform continuously on stage in a 10 meter by 20 meter world for one hour. The robot interacts with twelve other robots, stage elements, and human performers. Fast, accurate, and continuous state estimation for robot pose is a critical component for robots to safely perform on stage in front of a live audience. A custom robot platform was designed to use a Particle Filter to estimate state. The motor controller was developed to control robot vectoring and report odometry, and noise analysis on an absolute positioning system, Ubisense, was performed to characterize the system. High frequency noise confounds the Ubisense measurement of 0, but the Particle Filter acts as a low pass filter on the absolute positions and mixes the high frequency components of the odometry to determine an accurate estimate of the robot pose.