System identification of a free floating telerobot using Kalman filtering and a stereoscopic vision sensor

Due to the character of the original source materials and the nature of batch digitization, quality control issues may be present in this document. Please report any quality issues you encounter to digital@library.tamu.edu, referencing the URI of the item.Includes bibliographical references (leaves 78-79).Issued also on microfiche from Lange Micrographics.A telerobot has been acquired that floats on air bearings and is intended to simulate the dynamics of a spacecraft in a two-dimensional plane. The robot was delivered without a computer, sensors or documentation so an effort has been launched to determine how the apparatus worked and to identify the model parameters associated with mass, moment of inertia and thrust. The robot has been modified to accommodate a laptop as the onboard computer and a unique stereoscopic vision sensor as a navigation system. The unknown model parameters are then identified using both least squares estimation and Kalman filtering. The unique stereoscopic vision sensor system is based on one-dimensional position sensing diodes (PSD's) and active targets that broadcast a modulated signal. The active targets are mounted at known points in the robot frame of reference and broadcast their signal to the PSD sensors that are stationary in the inertial coordinate frame. This system enables real-time attitude measurements with no moving parts. The robot's mass, moment of inertia and the forces generated by its thrusters are identified using direct measurements and the well known linear least squares estimation algorithm. Identification using these techniques required experiments specifically designed to characterize the system. Some of the parameters may change over time, so a means of conducting on-line system identification was developed. A Kalman filter was designed which could simultaneously perform state estimation and parameter identification on-line. This technique did not require an experiment specifically designed for identification purposes and could accurately find the unknown model parameters during normal robot maneuvering