Compensating Path Dependent Nonlinearity in Precision Actuators

39 concepts27 papers1 proposal

In ideal mechanical systems, an input signal—such as voltage or pressure—produces a predictable, instantaneous output. However, in high-precision engineering and soft robotics, materials often exhibit hysteresis: a phenomenon where the state of an actuator depends not just on its current input, but on the entire history of its previous movements. This "memory" effect creates a nonlinear relationship that complicates control, leading to positioning errors and residual vibrations.

The Challenge of Material Memory

Hysteresis is most prevalent in "smart" materials like piezoelectric ceramics and compliant systems like pneumatic actuators. As Arachchige et al. note in their work on soft quadrupedal robots, pneumatic muscles exhibit a physical response that lags behind pressure inputs in a non-linear fashion. This lag is not a simple delay; it is a structural characteristic where the material follows different paths during loading and unloading.

In piezoelectric systems, this behavior is even more pronounced. Gu et al. describe this as a "memory effect" that prevents sub-micron accuracy in optical alignment. Because the actuator's displacement is path-dependent, a single voltage value could correspond to multiple possible positions depending on whether the voltage was recently increasing or decreasing. This uncertainty degrades the performance of sensitive instruments, such as the scanners used in atomic force microscopy Pota et al..

Modeling the Nonlinear Loop

To control a hysteretic system, researchers must first build a mathematical representation of this "loop." Several classical frameworks exist to approximate these dynamics:

  • The Duhem Model: Zhou utilizes this model to capture the behavior of piezoceramic actuators, using polynomial approximations to handle the complex differential equations that define the material's state transitions.
  • The Bouc-Wen Model: Often used in machining, this model captures the smooth transition between elastic and plastic-like states. Fu et al. applied this to 3D elliptical vibration cutting, using optimization algorithms to identify the specific parameters that define the actuator's unique "fingerprint."
  • The Preisach Model: This approach treats the actuator as a collection of many tiny "elementary operators," each with its own switching behavior. By summing these operators, Gu et al. were able to predict and compensate for the inherent memory of fiber-positioning units.

While these models are effective, they can be computationally expensive. Recent work by Chandra et al. suggests a shift toward Sparse Identification of Nonlinear Dynamics (SINDy). By using sparse regression, the authors discovered "white-box" mathematical models that are as accurate as deep learning but significantly faster to train and easier to interpret.

Linearization through Hardware Feedback

One of the most direct ways to overcome hysteresis is to "close the loop" using physical sensors. If the controller knows the actual displacement of the actuator in real-time, it can adjust the input to correct for the material's lag.

Bliss et al. demonstrated this at Lawrence Livermore National Laboratory by incorporating strain gauge feedback into piezoelectric mirrors. By measuring the actual physical strain, the system could ignore the nonlinear voltage-to-displacement relationship and focus on the desired position. Similarly, Imamura et al. developed "smart actuators" by bonding magnetostrictive films to piezoelectric materials. The film acts as a high-sensitivity sensor, providing the feedback necessary to linearize the actuator's movement, effectively canceling out the material's inherent non-linearity.

Data-Driven and Adaptive Control

In systems where physical sensors are too bulky or expensive, researchers turn to advanced control algorithms that "learn" the hysteresis.

For pneumatic systems, which are notoriously difficult to model due to friction and air compressibility, Groen et al. introduced feed-forward neural networks. These networks learn the complex dynamics of "rubbertuator" arms, allowing them to track trajectories with far greater precision than traditional PID controllers. Wu et al. followed a similar path for pneumatic cylinders, using data-driven approaches to regulate force despite nonlinear spring characteristics.

For high-speed applications, Smith proposed combining physics-based inverse models with L1 adaptive control. This strategy allows the system to adapt to changes in the material's behavior over time, ensuring that the actuator remains accurate even as it ages or operates under different environmental conditions. By predicting the error caused by hysteresis and applying a corrective "counter-signal," these controllers allow smart materials to meet the stringent requirements of modern precision engineering.

Key Foundational Work

<title>Real-time wavefront correction system using a zonal deformable mirror and a Hartmann sensor</title> (1991) — contributed 2 concept elements to this direction.
Analysis and control of a rubbertuator arm (1996) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_actuator_hysteresis_16f576ed,
  title = {actuator hysteresis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/actuator-hysteresis/rdr_16f576edf27545b5bd595fb6afae37fe},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-21},
}

This exact version (version 1, frozen text):

@misc{lacuna_actuator_hysteresis_16f576ed_v1,
  title = {actuator hysteresis},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/actuator-hysteresis/rdr_16f576edf27545b5bd595fb6afae37fe/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-21},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake