Developments in Adaptive Filtering and System Identification

Potential applications include optical communications, target tracking, and image processing.

Progress has been made on several fronts in a continuing program of research oriented toward the development of real-time algorithms, and computer and control systems that utilize the algorithms, for adaptive filtering, prediction, and system identification with improved efficiency and numerical stability in applications that involve large numbers of channels and high filter orders. Potential applications include adaptive optics, optical communications, target tracking, image processing, blind identification and deconvolution in wireless communications, and active control of noise and vibration.

The program has included three main subprograms supporting different aspects of research and development at the Air Force Research Laboratory on directed-energy weapons and laser communications:

  • Control (more precisely, suppression) of laser-beam jitter.

These Scoring-Camera Images, acquired in experiments in an adaptive-optics laboratory testbed, show that the laser beam was more tightly focused when adaptive control was used.
The most important sources of laser beam jitter are platform (aircraft) vibration and atmospheric turbulence. Jitter typically consists of multiple narrow frequency-band components, often combined with broad-frequency-band disturbances. The need for adaptive (in contradistinction to non-adaptive) control or suppression of jitter arises because the frequency content of jitter varies as different platform modes are excited and different atmospheric conditions are encountered. No linear time-invariant (LTI) controller can control all disturbance bandwidths optimally. By adapting to the particular frequency content of any disturbance, adaptive control effectively extends the bandwidth of even robust, high-performance LTI feedback controllers.

In this subprogram, lattice-filter- based subspace system- identification and adaptive control algorithms were applied in laser- beam-steering experiments. It was shown that enhanced rejection of disturbances is achievable in laser-beam steering by use of modern optimal feedback controllers augmented with adaptive control loops that determine control gains that are optimal for the current disturbance acting on the laser beam. In each adaptive control loop, an adaptive lattice filter implicitly identifies disturbance statistics from real-time data on the relative orientation of the laser beam as measured by use of a quad cell.

The most important recent advances have included the development of a method of variable-order adaptive control, a method of frequency weighting in adaptive control, and adaptive control of a new class of liquid-crystal beam-steering devices. Variable-order adaptive control is important because higher-order controllers are usually necessary for optimal steady-state performance but require more data for adaptive identification of optimal gains and, hence, produce slower adaptation and, if adaptive loops are closed before nearly optimal gains are identified, often produce large transient responses. Since the fewer gains needed for lower order control laws can be identified from fewer data points, lower order control laws yield faster adaptation without generating large transients.

  • Using new liquid-crystal devices to control laser beams.

Feedback and adaptive feedforward control have been applied to a family of liquid-crystal laser-beam-control devices being developed as actuators in jitter control and adaptive optics for applications to laser weapons and laser communications. In comparison with fast-steering-mirror laser-beam-control devices, the liquid-crystal devices offer the advantages of low power consumption and no moving parts. It was necessary to modify a prior adaptive control loop to accommodate nonlinearities in the liquid-crystal devices resulting from rate-limit and quantization effects not encountered in fast steering mirrors.

  • Adaptive control and filtering in adaptive optics.

The classical adaptive-optics control loop in a prior adaptive-optics laboratory testbed used in development of directed-energy weapons was augmented with an adaptive control loop to enhance beam control and imaging in the presence of atmospheric turbulence. High-fidelity wave optics simulations predicted significant improvements in the Strehl ratio (a measure of on-target beam intensity) and tracking jitter. Improvements similar to those predicted have been confirmed by results of experiments: In the experiments, the first 150 modes from a set of frequency-weighted deformable mirror modes were used by the adaptive control loop. First, 3,000 wavefront-sensor frames were used to identify the adaptive filter gains, and then the performance of the adaptive controller was evaluated on 1,000 frames independent of those used for identification. For comparison, the same experiment was performed with only the classical adaptive-optics and tracking loops, using the same 1,000 frames for evaluation. For the turbulence scenario examined, the use of the adaptive controller was found to increase the Strehl ratio by nearly 50 percent and reduce the variability by more than 15 percent. The figure shows example images, acquired during the evaluation sequences, that illustrated the benefit afforded by the adaptive control loop.

This work was done by Steve Gibson, Nestor Perez, Pawel Orzechowski, Neil Chen, and T. C. Tsao of the University of California; Darryl Sanchez, Troy Rhoadarmer, Laura Klein, and Robert Vincent of the Air Force Research Laboratory; and Bruce Winker, Milind Mahajan, and Bing Wen of Teledyne Scientific Co. for the Office of Naval Research.


This Brief includes a Technical Support Package (TSP).
Document cover
Developments in Adaptive Filtering and System Identification

(reference ONR-0001) is currently available for download from the TSP library.

Don't have an account? Sign up here.