Vision is a flexible, non-contact sensor that can be used for position feedback in
closed-loop control of dynamic systems. Current vision systems for industrial
automation provide low sample rates and large sample delays relative to other types of
position sensors. Poor sample rates and sample delays are a result of the vast volume of
data that must be collected and processed by the vision system. A predictive visual
tracker can help compensate for some of the deficiencies of current industrial vision
systems. The objectives of the present research are to demonstrate that vision is a useful
feedback sensor and prediction can be used to improve performance by compensating for
the feedback delay of the vision system.
An inverted pendulum was stabilized using a vision sensor as feedback to a state-feedback
controller. The vision data was run through a d-step ahead predictor to
compensate for the vision system delays. The system was simulated in Mat lab and an
actual physical system was used to test the performance of the control system.
The inverted pendulum provides a good test-bed for studying predictive control
using vision feedback. The pendulum will fall without the constant adjustment of the
cart position. The adjustment of the cart by the controller is delayed because of latency
and quantization errors in vision feedback. The better the controller is able to
compensate for delays and quantization errors, the greater its ability to stabilize the
inverted pendulum. / Graduation date: 1996
| Identifer | oai:union.ndltd.org:ORGSU/oai:ir.library.oregonstate.edu:1957/34657 |
| Date | 17 August 1995 |
| Creators | Barrett, Spencer Brown |
| Contributors | Kolodziej, Wojtek |
| Source Sets | Oregon State University |
| Language | en_US |
| Detected Language | English |
| Type | Thesis/Dissertation |
Page generated in 0.0018 seconds