Complex Stiffness Model of Physical Human-Robot Interaction: Implications for Control of Performance Augmentation Exoskeletons

Binghan He, Huang Huang, Gray C. Thomas, Luis Sentis

TL;DR

This paper introduces a "complex stiffness" model for human joints that includes hysteretic damping, better explaining observed human-exoskeleton interaction dynamics than traditional models. This new model allows for more robust and performant exoskeleton control by accounting for this inherent human damping.

What Problem Does This Paper Address?

Designing exoskeletons that seamlessly augment human strength (performance augmentation exoskeletons) is challenging because the human and robot are interconnected. The stability of such systems heavily relies on accurately modeling the human's dynamic behavior. Traditional models often use a simple mass-spring-damper system, but these fail to capture key observed phenomena like a consistent damping ratio across varying stiffness and inertia, and a non-zero phase shift at low frequencies. This gap in modeling accuracy can lead to overly conservative control strategies or instability.

Methods

The researchers propose and compare three models for human elbow joint dynamics when interacting with an exoskeleton:

  1. Linear Mass-Spring-Viscous Damper: A standard model.
  2. Complex Stiffness (Hysteretic Damping): A model incorporating a real stiffness and a hysteretic damping term, represented in the frequency domain.
  3. Combined Model: A model including real stiffness, viscous damping, and hysteretic damping.

They conducted experiments using an elbow-joint exoskeleton on a human subject. By varying exoskeleton inertia and human joint stiffness (through bias torque and grip force), they collected frequency domain data. Statistical F-tests were used to compare the accuracy of the three models, with a focus on the significance of the hysteretic damping term. They also explored a linear relationship between the hysteretic damping and real stiffness, leading to a simplified 1-parameter model. Finally, they designed a fractional-order controller to leverage this hysteretic damping for improved exoskeleton performance.

Datasets

Not discussed in the paper. The data used was collected from experiments with a single human subject using an elbow-joint exoskeleton.

Key Insights and Ideas

  • Hysteretic Damping is Key: The most significant finding is that human joints exhibit hysteretic damping, not just viscous damping. This is crucial for accurately modeling their dynamic behavior.
  • Explaining the "Unexplained": The complex stiffness model, specifically the hysteretic damping term, successfully explains the observed low-frequency phase lag and the near-constant damping ratio of the human-exoskeleton system, which traditional models struggle with.
  • Linear Relationship Simplifies Things: A surprising and useful discovery is the linear relationship between the hysteretic damping coefficient ($C_h$) and the real stiffness ($K_h$). This allows for a significant simplification of the model to a single parameter, making controller design more tractable.
  • Fractional-Order Control Potential: The hysteretic damping can be exploited by advanced control techniques like fractional-order controllers to enhance performance and stability in augmentation exoskeletons.

Key Results

  • Improved Model Accuracy: Statistical F-tests showed that the hysteretic damping term ($C_h$) significantly improved model accuracy compared to models without it (Model 3 vs. Model 1). The contribution of hysteretic damping was found to be more significant than viscous damping.
  • Consistent Damping Ratio: The complex stiffness model accurately predicts the near-constant damping ratio observed across varying stiffness and inertia, a phenomenon not explained by simpler models.
  • Linear $C_h$-$K_h$ Relationship: A strong linear relationship was identified between the hysteretic damping ($C_h$) and real stiffness ($K_h$), with a high coefficient of determination ($R^2 = 0.85$ for Model 2). This enabled a simplified 1-parameter model.
  • Controller Design Implications: The findings suggest that controllers can be designed to exploit this hysteretic damping for improved strength amplification while maintaining stability.

Limitations and Future Work

  • Single Subject: The study was conducted with a single subject. While the results are compelling, further studies with more subjects are needed to confirm the universality of the observed linear relationship and to establish population variance for model parameters.
  • Amplitude Dependence: The paper notes that the hysteretic damping might be amplitude-dependent, similar to coulomb friction. Future work could investigate this relationship and its implications for controller robustness across different force inputs.
  • Controller Tuning: While fractional-order controllers offer advantages, they still require hand-tuning for different users. Future research could focus on automated tuning methods.
  • Passive Hysteresis: The authors suspect the hysteresis is a passive material property from muscle contraction. Further investigation into the physical origins of this damping would be valuable.

Summary updated

Abstract

Human joint dynamic stiffness plays an important role in the stability of performance augmentation exoskeletons. In this paper, we consider a new frequency domain model of the human joint dynamics which features a complex value stiffness. This complex stiffness consists of a real stiffness and a hysteretic damping. We use it to explain the dynamic behaviors of the human connected to the exoskeleton, in particular the observed non-zero low frequency phase shift and the near constant damping ratio of the resonance as stiffness and inertia vary. We validate this concept with an elbow-joint exoskeleton testbed (attached to a subject) by experimentally varying joint stiffness behavior, exoskeleton inertia, and the strength augmentation gain. We compare three different models of elbow-joint dynamic stiffness: a model with real stiffness, viscous damping and inertia; a model with complex stiffness and inertia; and a model combining the previous two models. Our results show that the hysteretic damping term improves modeling accuracy (via a statistical F-test). Moreover, this term contributes more to model accuracy than the viscous damping term. In addition, we experimentally observe a linear relationship between the hysteretic damping and the real part of the stiffness which allows us to simplify the complex stiffness model down to a 1-parameter system. Ultimately, we design a fractional order controller to demonstrate how human hysteretic damping behavior can be exploited to improve strength amplification performance while maintaining stability.
BibTeX
@inproceedings{he2019complex,
  author = {Binghan He and Huang Huang and Gray C. Thomas and Luis Sentis},
  title = {{Complex Stiffness Model of Physical Human-Robot Interaction: Implications for Control of Performance Augmentation Exoskeletons}},
  booktitle = {IEEE/RSJ International Conference on Intelligent Robots and Systems},
  year = {2019},
  url = {https://openreview.net/forum?id=5n8tK1Rd0j},
}
Figure