Passivity-Guaranteed Variable Impedance and Compliance Control via Online Parameter Optimization

Huseyin Tugcan Dinc, Ersin Daş, Nikolas Thelenberg, Christian OttarXiv 2026 · arXiv

This paper introduces a novel framework for guaranteeing the stability—specifically, the passivity—of robots that need to change their stiffness and inertia on the fly. By solving a convex quadratic program (QP) in real-time to modulate inertia scaling, the proposed method enables aggressive impedance adaptation without the performance trade-offs or "energy budget" limits that plague traditional approaches.

What Problem Does This Paper Address?

Variable Impedance Control (VIC) is a powerful tool in robotics, allowing a manipulator to be "soft" when interacting with humans and "stiff" when performing high-precision tasks like assembly or cutting. However, changing these parameters online—especially rapid increases in stiffness—can inadvertently "inject energy" into the control loop.

In the world of control theory, this can lead to a violation of passivity, meaning the robot might start generating its own energy, leading to dangerous oscillations or instability when it hits an object or is pushed by a human. Existing solutions usually involve adding massive amounts of damping (which makes the robot sluggish) or using "energy tanks"—virtual reservoirs that store dissipated energy to "pay" for future stiffness increases. The problem with tanks is that they can run dry; once depleted, the robot is forced to stop adapting its stiffness, potentially failing its task.

Methods

The authors propose an online parameter optimization approach that replaces the "virtual tank" with a real-time mathematical constraint.

  1. Optimization-based Passivation: At every control cycle (1 kHz), the system solves a Quadratic Program. The objective is to stay as close as possible to the user's desired stiffness and inertia while satisfying a state-dependent passivity constraint.
  2. Inertia Modulation: The core innovation is treating the desired inertia as a controllable scaling factor ($\epsilon$). When a task demands a sharp increase in stiffness (which injects energy), the optimizer automatically reduces the inertia scaling to compensate, keeping the total energy balance passive.
  3. Compliance Control via Eigendecomposition: For complex scenarios where the robot's inertia changes based on its posture, the authors use a Cholesky-based eigendecomposition. This allows them to optimize the "spectral weights" (eigenvalues) of the stiffness matrix, ensuring the robot's behavior remains physically consistent in all directions.
  4. Critical Damping: The framework is designed to maintain a critically damped response throughout the adaptation, ensuring the robot reaches its target state as quickly as possible without overshooting.
Figure 1: Overview of the proposed variable impedance and compliance control framework.

Datasets

The method was validated through physical experiments on a Kinova Gen3 7-DoF manipulator equipped with a Bota SensONE 6-axis force/torque sensor. Rather than using a static dataset, the researchers tested the controller against:

  • Sinusoidal Stiffness Profiles: Smoothly varying the robot's "feel" over time.
  • Pulse-Wave Profiles: Testing the limits with near-instantaneous, aggressive jumps in stiffness.
  • Human Perturbations: A human operator physically interacted with the robot to inject external energy and test the controller's ability to maintain stability.

Key Insights and Ideas

  • Inertia as a Dynamic Buffer: Traditionally, inertia is viewed as a fixed property of the controller. Here, it acts as a "shock absorber" for the energy balance. By modulating inertia scaling, the optimizer can "hide" the energy injected by stiffness changes.
  • Reducing Conservatism: Previous stability proofs were often "state-independent," meaning they assumed the worst-case scenario and were overly restrictive. This method is "state-dependent"—it only restricts the robot as much as the current motion requires, unlocking a much wider range of usable stiffness.
  • Intrinsic Storage vs. Auxiliary Tanks: By using the storage function of the impedance dynamics itself, the method avoids the "empty tank" problem. It doesn't need a history of dissipated energy to allow for a future stiffness increase; it simply balances the energy equation in real-time.

Key Results

  • Superior Tracking: The robot successfully tracked "pulse-wave" stiffness transitions—nearly vertical steps in stiffness—that would typically destabilize a standard VIC controller.
  • No Depletion: In head-to-head comparisons with energy tank methods, the tank-based controller failed to track the desired stiffness once its energy budget was depleted (around 15 seconds into the test). The proposed QP method tracked the full profile indefinitely.
  • Real-Time Feasibility: Despite the complexity of solving a QP and performing eigendecompositions, the entire loop ran reliably at 1 kHz on a standard desktop processor.
Block diagram representation of the proposed (a) variable impedance controller and (b) compliance controller.

Limitations and Future Work

The authors acknowledge a few areas for improvement:

  • Numerical Robustness: The eigendecomposition can become sensitive if the robot's eigenvalues become very close to each other (repeated eigenvalues). Future work will investigate "orthogonal subspace alignment" to fix this.
  • Prediction Horizon: The current setup is a "one-step" optimizer. Extending this to a multi-step Model Predictive Control (MPC) horizon could allow the robot to anticipate future stiffness needs and prepare the energy balance more smoothly.
  • Subspace Alignment: Investigating better ways to handle the numerical consistency of eigenvectors during rapid configuration changes.

Summary updated

Abstract

Variable impedance control enhances the interaction capability of robot manipulators by adapting impedance parameters online. However, time-varying impedance parameters may inject energy into the system and violate passivity, particularly when tasks demand rapid stiffness changes. Existing methods address this by increasing damping or limiting the stiffness profile, which often degrades the dynamic response. In this work, we propose a quadratic programming-based parameter optimization framework that guarantees passivity by modulating the desired inertia to compensate for energy injected by stiffness variations, while preserving a critically damped system. For the case of diagonal impedance matrices, we optimize the rates of change of inertia and stiffness subject to a passivity constraint formulated as a convex quadratic program. We extend the framework to compliance control, where the desired inertia is coupled to the robot's configuration-dependent inertia matrix. We use an eigendecomposition based on Cholesky factorization to optimize the spectral weights while maintaining passivity. Experiments on the Kinova Gen3 7-degree-of-freedom manipulator validate both methods with human interaction perturbations. Comparisons with an energy tank approach and a nonpassivity-based method demonstrate that the proposed framework achieves less conservative stiffness tracking without requiring an energy reservoir or damping overhead.
BibTeX
@misc{dinc2026passivity,
  author = {Huseyin Tugcan Dinc and Ersin Daş and Nikolas Thelenberg and Christian Ott},
  title = {{Passivity-Guaranteed Variable Impedance and Compliance Control via Online Parameter Optimization}},
  year = {2026},
  eprint = {2611.10778},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.10778},
}
Figure