TL;DR
This paper introduces the F-HASEL, a novel soft actuator design that decouples sensing from actuation using an additional electrode pair. This innovation, combined with miniaturized circuitry, enables accurate displacement estimation at high frequencies (up to 20 Hz) and with compact power supplies, paving the way for more responsive and untethered soft robotic systems.
What Problem Does This Paper Address?
Soft actuators, particularly Hydraulically Amplified Self-Healing Electrostatic (HASEL) actuators, are highly desirable for applications requiring compliance and adaptability, such as human-robot interaction, prosthetics, and biomimetic robots. A key challenge in realizing their full potential is accurate proprioception – knowing the actuator's exact position or displacement without external sensors. While capacitive self-sensing has shown promise for HASEL actuators, existing methods struggle with high actuation frequencies (above 1 Hz) and require bulky, expensive high-voltage (HV) components, limiting their use in portable and untethered applications. This paper tackles the limitations of existing capacitive self-sensing techniques for HASEL actuators, specifically addressing the need for high-frequency operation and miniaturization.
Methods
The core innovation is the F-HASEL (Frog-HASEL) actuator. Unlike traditional Peano-HASEL actuators where the same electrodes are used for both actuation (high voltage) and sensing (low voltage signal superimposed), the F-HASEL introduces a separate, dedicated pair of electrodes solely for sensing. These "LV sensing electrodes" are positioned on the opposite side of the actuator from the HV actuation electrodes.
This separation allows for two primary sensing methods:
- Voltage-Method: This method treats the LV sensing electrodes and a series resistor as a passive first-order low-pass filter. Changes in the actuator's capacitance (due to deformation) shift the filter's cutoff frequency, which in turn alters the measured output voltage. This voltage change is then mapped to displacement.
- Impedance-Method: This approach uses both voltage and current measurements across the LV sensing electrodes to calculate the complex impedance. Changes in impedance, directly related to capacitance and thus displacement, are then used for estimation.
Crucially, the sensing circuitry is designed to be low-voltage (LV) and uses off-the-shelf components, significantly reducing cost and form factor. The sensing signal path is entirely separate from the HV actuation path, making the sensing method polarity-agnostic, which helps mitigate charge retention issues in the actuator. For multi-actuator applications, a multiplexing strategy is employed to allow a single sensing circuit to sequentially measure multiple F-HASELs.
Datasets
Not discussed in the paper. The experiments rely on ground truth measurements from a laser sensor and custom setups for evaluating actuator performance.
Key Insights and Ideas
- Decoupling Sensing and Actuation is Key: The F-HASEL's dedicated sensing electrodes are the fundamental breakthrough, resolving the interference and limitations inherent in superimposing signals on the same electrodes.
- High-Frequency Sensing is Achievable: By separating the sensing electrodes, the authors demonstrate accurate displacement estimation at frequencies up to 20 Hz, a significant leap from the 1 Hz limitations of prior work.
- Miniaturization and Portability: The use of LV, off-the-shelf components for the sensing circuitry drastically reduces size and cost, enabling truly portable and untethered soft robotic applications.
- Polarity Agnosticism Matters: The ability to reverse HV polarity is crucial for mitigating charge retention, allowing actuators to return to their initial state more reliably.
- Hysteresis Requires Advanced Mapping: At higher frequencies, actuator hysteresis becomes a significant factor. The paper shows that a simple mapping isn't enough, and a "dual third-order polynomial mapping" that accounts for contraction vs. relaxation is necessary for accurate estimation.
Key Results
- High-Frequency Performance: The F-HASEL with the proposed sensing circuitry achieved accurate displacement estimation up to 20 Hz, with a significant reduction in Normalized Root Mean Squared Error (NRMSE) compared to prior state-of-the-art methods (e.g., NRMSE reduced from ~0.1 to ~0.03 at 1 Hz, and from ~0.8 to ~0.05 at 5 Hz).
- Reduced Phase Lag: The phase lag between estimated and ground truth displacement was drastically reduced, staying under 4 degrees even at higher frequencies, compared to ~32 degrees in previous work.
- External Load Tracking: The system successfully tracked displacement changes caused by both actuation voltage and applied external tensile forces, demonstrating robustness.
- Wearable VR Application: A wearable system was demonstrated, successfully tracking knee and hip joint rotations of a VR user in real-time using four F-HASEL actuators.
- Noise Reduction: The design and instrumentation amplifier choice led to a significant reduction in noise on the sensing signal compared to previous methods.
Limitations and Future Work
- Update Rate for Multiple Actuators: The multiplexing approach for multiple actuators limits the update rate for each individual actuator, which might be a bottleneck for applications requiring very fast, simultaneous tracking of many joints.
- LV Sensing Electrode Geometry Optimization: The authors suggest that optimizing the geometry, size, and position of the LV sensing electrodes could further enhance sensing performance.
- Parasitic Coupling Noise Mitigation: While noise was reduced, parasitic coupling remains a source of interference. Future work could focus on improving component placement and isolation to further minimize this noise.
- Hysteresis Mitigation: While the dual polynomial mapping addresses hysteresis, further exploration into actuator designs or control strategies that inherently reduce hysteresis could be beneficial.
Summary updated