Robust Multivariable Feedback Design for Uncertain Linear Systems
Abstract
Realistic control strategies must address the inevitable uncertainty which accompanies nominal system descriptions. Even though uncertainty can often be characterized mathematically, effective robust control techniques have been slow to appear. This work investigates robust control methods directed at both the analysis and design of multivariable feedback systems in the presence of system uncertainty. The first part of the thesis examines perturbed interaction from the generalized Nyquist/characteristic locus perspective. This work establishes that, for a given class of uncertainty and a specific class of gain- limited controllers, feedback compensation can be optimally deployed to reduce perturbed interaction. Subsequently, this treatment exploits the geometric eigen-structure embodied in the characteristic locus framework along with the appropriate stationary conditions in order to characterize the worst case uncertainty which produces the largest interaction as measured by the perturbed misalignment angles. Furthermore. the structure of the worst case perturbation has a particularly simple representation which facilitates the determination of the worst case interaction based on simple open-loop quantities. These results together with the previous development of the E-Contour method complete the overall development of the characteristic locus approach as a convenient robust analysis tool.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA258257
Entities
People
- Mark C. Crews
Organizations
- Air Force Institute of Technology