Achieving Dexterous Bidirectional Interaction in Uncertain Conditions for Medical Robotics

Carlo Tiseo, Quentin Rouxel, Martin Asenov, Keyhan Kouhkiloui Babarahmati, Subramanian Ramamoorthy, Zhibin Li, Michael N. MistryarXiv 2022 · DOI · arXiv

Abstract

This paper introduces a versatile control architecture for medical robotics designed to operate across diverse clinical scenarios without task-specific manual tuning. Utilizing Fractal Impedance Control (FIC), the system ensures passivity-based stability, making it robust against communication latencies and the non-linear mechanics of biological tissues. We demonstrate the architecture's efficacy through a series of experiments involving robotic surgery, physical rehabilitation, and diagnostic ultrasound. Our results indicate that a single set of control parameters can successfully manage varied interaction requirements, from high-precision cutting to compliant patient assistance.

Introduction

The field of medical robotics is currently characterized by task-specific fragmentation. Robots designed for surgical precision operate under fundamentally different control paradigms than those intended for stroke rehabilitation or diagnostic imaging. Traditional interaction controllers often rely on precise contact models, which are difficult to maintain when interacting with "mushy," non-linear biological tissues. Furthermore, teleoperation introduces significant challenges, including communication latency—which can induce dangerous oscillations—and the need for frequent manual gain adjustments when switching between admittance (following) and impedance (resisting) behaviors. This work proposes a modular, FIC-based architecture to address these "uncertain conditions" autonomously and safely.

Methodology

The proposed system utilizes a master-replica configuration featuring a Sigma.7 haptic interface and 7-DOF Franka Emika Panda arms. The core of the approach is a control stack with several key components:

  • Fractal Impedance Control (FIC): A passivity-based controller that is inherently robust to time delays. Its passive nature allows multiple control layers to be superimposed without compromising overall system stability.
  • SEIKO Retargeting: An algorithm using Sequential Quadratic Programming (S-QP) that runs at 1 kHz to ensure movements stay within physical joint and torque limits while avoiding kinematic singularities.
  • NonLinear-PD (NLPD): Used to provide haptic feedback to the operator, mimicking the forces experienced by the replica robot.
The integrated control architecture showing the Master-Replica loop and the FIC-based feedback mechanisms across surgery, rehabilitation, and diagnostic tasks.

Experimental Results

The architecture was evaluated across four distinct physical scenarios: scalpel surgery on a silicone skin model, a rehabilitation brace setup, diagnostic ultrasound scanning of a gelatin phantom, and bimanual handling of a fragile potato chip using soft optical force sensors (TACTIP).

Experimental setups for scalpel surgery and rehabilitation, illustrating the operator's perspective and the robotic interaction.

A key finding was the "Zero-Tuning" capability: identical controller gains were used for all tasks. In the surgery task, the operator completed 16 successful incisions. In rehabilitation, the system transitioned seamlessly between a "transparent" mode (following the patient with ~8N force) and an "assistance" mode (applying up to 40N force) without stability loss. The ultrasound experiment successfully identified internal masses, with the FIC dynamically adjusting stiffness to maintain contact with irregular surfaces.

Discussion and Future Work

While the control architecture demonstrated high robustness, the study identified human-centric bottlenecks. Operators faced challenges with 3D depth perception using 2D camera feeds, leading to non-linear cutting paths. Additionally, the inertia mismatch between the lightweight master device and the heavier replica arm caused operator fatigue during extended tasks. Future research will focus on integrating Augmented Reality (AR) for better spatial awareness and developing haptic cues to communicate the robot's residual range of motion.

Conclusion

The FIC-based architecture represents a significant step toward "plug-and-play" medical robotics. By decoupling stability from specific task parameters, this approach enables a single robotic platform to serve multiple clinical roles, reducing the technical burden on medical personnel.

Summary updated

BibTeX
@misc{tiseo2022achieving,
  author = {Carlo Tiseo and Quentin Rouxel and Martin Asenov and Keyhan Kouhkiloui Babarahmati and Subramanian Ramamoorthy and Zhibin Li and Michael N. Mistry},
  title = {{Achieving Dexterous Bidirectional Interaction in Uncertain Conditions for Medical Robotics}},
  year = {2022},
  eprint = {2206.09906},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2206.09906},
}
Figure