September 2005
3D TMM visulization movies

April 2004
research status poster (348 kb pdf file)

December 2003

Outline:
Research Status as of April 2003
As of April 2003 I had completed upgrading the computer and the real time operating system used to control the robot I work on known as SAMII (if you haven't already done so, you can read about SAMII in the introduction to my research).  I was attempting to recreate a mass damping controller for SAMII similar to that implemented by previous researchers.  I had not done anything intentionally different from the previous researchers, but my controller was unstable.  Here is a video clip (1mb Quicktime .mov file) of SAMII after the unstable controler has been turned on.  (If you have problems viewing the video in your browser, right click and choose Save Target As to download the file).  (Additional movie clips can be found here).
(return to outline)

Potential Causes for Controller Instability
In comparing the actual system to the controller development in Lynnane George's thesis, there were two differences between the actual system and the model that were suspected as causes of the controller instability.  The first is that the mass damping voltage should be proportional to the position of the flexible base while the actual system uses an accelerometer.  The second is that the hydraulic actuator does not act as a velocity source near the second natural frequency of the flexible base due to interaction with the base.  We considered using strain gauges to sense the position of the flexible base.  A neural network controller and velocity feedback control were considered as ways to force the actuator to act as a velocity source.  Some way to determine the viability of these two options was sought.
(return to outline)

System Identification
Experimental identification of SAMII and his flexible base operating about a nominal configuration was done in an attempt to better understand why the system was unstable.  It was also hoped that the experimentally determined system model might provided some insight into the effectiveness of the modifications to the system and the controller that were being considered (i.e. installing strain gauges and developing a controller that made the actuator act like a velocity source).
more on system identification
(return to outline)

Root Locus Analysis
My system identification efforts resulted in a linearized model for the system that could be used for root locus analysis.  Not only did this analysis provide insight into the cause of the instability of the mass damping controller, it also provided a way to simulate the effects of modifying the system.
more on root locus analysis
(return to outline)

Pole Placement Controller Design
Based on the linear model found in my system identification work, a sixth order state-space model that included the first two modes of the flexible base was developed.  Based on this state-space model, a pole placement controller was designed to place the poles associated with the first two modes of the flexible base at locations with damping of 0.7 (without active vibration suppression, the damping of these two modes is approximately 0.05 and a single-input-single-output controller with a low-pass filter on the accelerometer signal is able to increase the damping of the first mode to approximately 0.25).

This pole-placement controller looked very promising in simulation, but in practice it made the third mode of the flexible base unstable.  This lead to curve fitting a wider frequency range of the Bode data to include the third mode in my models.
(return to outline)

System Identification with 3 Modes
After implementing a pole placement controller controller that placed the poles for the first two modes of SAMII's flexible base and realizing that this controller made the third mode unstable, I set out to curve fit a wider frequency range of my data to get a model that included the third mode of the flexible base.
more about system i.d. with 3 modes
(return to outline)

Pole Canceling Compensator Design
Once I had a model that included the first three flexible base modes, I designed a compensator that canceled the three lightly modes and replaced them with critically damped ones.  I then used a root locus to chose the overall gain for the compensator so that the damping for the closed loop poles was 0.7.  This controller worked well in open loop essentially shaping the input so that base vibration was not excited.  However, because  of the small overall gain for the compensator, this control scheme was not effective and rejecting vibration disturbances.
more on design and implementation of a pole canceling compensator
(return to outline)

Additional Movie Clips
Links to additional movie clips including a side-by-side comparison of SAMII with and without mass damping control can be found on this page: movie_clips.html.
(return to outline)

junk


 
Mechanical Engineering

Georgia Tech

Dr. Wayne Book

Intelligent Machine Dynamics Laboratory

Last Updated on December 15, 2003