Office of Technology Transfer – University of Michigan

Digital Adaptive Control Algorithm Based on a Retrospective Correction Feedback Filter

Technology #4048

UM File # 4048

Background
One of the important challenges faced in control system design is the minimization of external disturbance signals. For applications such as active noise and vibration control, adaptive methods are useful for the solution of this most crucial problem. However, conventional methods in this field either require a direct measurement or a statistical description of the disturbance. This causes poor performance and instability due to modeling requirements where the system is time varying or difficult to identify. Thus, it is desirable to come up with adaptive control algorithms that apply to a large class of systems under minimal modeling requirements. With the potential to impact applications in various markets such as UAV ( > $4.4B ) and engine control technology ( > $900M ), active adaptive noise and vibration control comprises a promising area for research.

Technology Description
A discrete-time adaptive control algorithm has been developed at the University of Michigan for noise and vibration suppression. The strength of the proposed method comes from its estimation of the current controller’s performance based upon prior data. Most of the existing methods are either limited to tonal disturbances or require a direct measurement of the disturbance signal. By evaluating performance in terms of past data, the technique is “retrospective” in character; making it superior due to its minimal modeling requirements. Proven by numerical investigations and experiments on an aircraft fuselage, features such as tunable adaptation rate and robustness under uncertainty allow the algorithm to be highly effective and applicable to adaptive command following, disturbance rejection, stabilization and model reference control.
Applications • Unmanned Aerial Vehicles (UAVs) • Engine control • Noise cancellation • Noise and vibration control
Advantages • No direct measurement of the disturbance signal • Minimal modeling requirements • Greater robustness to noise and parameter uncertainty • Effective for unstable, MIMO and/or non-minimum phase systems