Calibration of a Four Degree of Freedom Stewart Platform Sensor Subject to Translational and Rotational Constraints

Open Access
Author:
Winslow, Anna Tatum
Graduate Program:
Aerospace Engineering
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
December 05, 2012
Committee Members:
  • Edward A Smith, Thesis Advisor
  • George A Lesieutre, Thesis Advisor
Keywords:
  • Stewart platform
  • sensor
  • Gough platform
  • orientation
  • calibration
Abstract:
Stewart platforms are typically fitted with linear actuator legs which serve to drive the mobile platform to some desired orientation and position. In more recent work, the Stewart platform has been fitted with sensor, rather than actuator, legs so that the platform could be used to quantify applied forces and torques. The orientation of the platform is predicted using the embedded sensor measurements and positions; therefore, it is necessary to calibrate the platform by finding updated positions of the sensors so that the accuracy of the predicted platform orientation and position are improved. A constrained four degree of freedom Stewart platform with four embedded linear displacement sensors was calibrated. In the case of the platform considered, the goal was to accurately determine the orientation of the platform based on the measured changes in the linear displacement sensor leg lengths. An existing test fixture was used to drive the platform to known static angles about three axes, to a maximum of approximately 10°. The test fixture applied pure rotations to the mobile plate of the platform and did not permit the mobile platform to translate. An orientation sensor, which incorporated a dual-axis inclinometer and a rotary encoder, was designed to interface with the test fixture to directly and accurately measure the orientation of the platform. The modular orientation sensor was designed to be easy to install and remove. Furthermore, the attachment plate of the orientation sensor was given a simple design so that the sensor could be used to calibrate nearly any Stewart platform. A calibration algorithm was developed which minimized the error of the orientation angles calculated from the linear displacement sensor measurements by updating the positions of the embedded displacement sensors relative to their as-designed values. The calibration was shown to reduce the maximum error in the predicted orientation angles by a factor of 10 to 0.5° or less. The average errors in the predicted orientations were all less than 0.1° after calibration.