Adaptive Nonlinear Autopilot for Anti-Air Missiles.
Abstract
The prevalent method of synthesizing nonlinear missile autopilots is by gain-scheduling linear designs. Although this approach has proven successful in numerous applications, the desire to continually improve performance without incurring additional cost suggests the need for a new design paradigm. An opportunity to address this need has been identified from previous research which employed neural network technology to augment approximate dynamic inversion controllers. In the one architecture a neural network adaptively cancels linearization errors through on-line learning, which may be accomplished by a weight update rule derived from Lyapunov theory. This effectively guarantees stability of the closed-loop system. This paper concerns a similar implementation in which neural networks function instead to improve command tracking of gain-scheduled control laws. This theoretical development is then specialized to the problem of synthesizing a bank-to-turn autopilot for an agile anti-air missile. Finally, the resulting hybrid control law is demonstrated in a nonlinear simulation and its performance is evaluated relative to that of the unaugmented gain-scheduled autopilot.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1998
- Accession Number
- ADA356503
Entities
People
- Donald T. Stansbery
- Michael B. Mcfarland
Organizations
- Air Force Research Laboratory