PaperlayerView exact revision on arXiv

Source-linked AI summary

A Physics-Based Closed-Loop Robotic Bioprinting Framework Towards Volumetric Muscle Loss Treatment

Omid Rezayof, Jerin T. Andrews, Ehsan Zobeidi, Ali Ghasemkhani, Meenakshi Kamaraj, Maryam Tilton, Johnson V. John, Farshid Alambeigi

arXiv:2609.12159v1cs.RO

TL;DR

For VML, DIW parameter selection matters, but existing approaches can require resource-intensive experiments, precise material characterization, or large datasets. This paper presents a physics-based closed-loop system that uses vision-based thickness feedback to adapt printing pressure. Across the reported tests, it converged below 0.5 mm thickness error, while open-loop errors reached 5 mm.

  • Problem

    For VML, DIW parameter selection matters, yet DoE is time- and material-intensive and alternatives can require precise material characterization or large datasets.

    Source: §I · p. 1

  • Method

    The framework uses autonomous vision-based thickness evaluation as feedback for closed-loop printing-pressure control.

    Source: B. Direct Control of Printing Pressure · p. 3, §IV · p. 8

  • Results

    Thickness errors fell below 0.5 mm with the controller on, versus as high as 5 mm in open-loop tests; convergence averaged 5.2 seconds.

    Source: §III · p. 7, §IV · p. 8

  • Takeaways & Limitations

    The framework offers an alternative to material- and time-intensive DoE parameter selection through closed-loop vision feedback.

    Source: §IV · pp. 7–8

Abstract

from arXiv · show

Robotic bioprinting and Direct Ink Writing (DIW) are being explored towards the treatment of Volumetric Muscle Loss (VML). While previous studies have shown the importance of proper parameter selection on the print outcome, existing approaches often rely on time- and material-intensive design of experiments methods, or require large, well-curated datasets for training machine learning models. In this paper, we propose a physics-based closed-loop robotic bioprinting system capable of near real-time parameter adaptation. The system integrates a 3D point cloud camera and fully autonomous vision-based algorithms to provide quantitative evaluation of printed constructs. This evaluation is fed into a controller that adjusts printing parameters to achieve a desired bead thickness. To assess the framework's performance, four experimental configurations were tested, each repeated three times. In these tests, printing began from an arbitrary initial parameter value, and the controller was tasked with adjusting the parameters to reach the desired thickness. The system converged in all trials, achieving a tracking error below 0.5 mm within an average of 5.2 seconds from the start of printing. The low standard deviation of the converged pressure over different tests (0.04 bar on average) demonstrates robustness and repeatability. Additional experiments were conducted with the controller turned off, enabling direct comparison with open-loop DIW bioprinting, further confirming the effectiveness of the proposed closed-loop framework in achieving the desired bead geometry.

I. INTRODUCTION

VML can exceed natural muscle regeneration, while DIW outcomes depend on printing parameters often tuned through resource-intensive or limited approaches. The paper presents a physics-based, vision-guided closed-loop framework that adjusts printing parameters to achieve desired bead thickness.

I. INTRODUCTION

  • VML injuries can exceed natural muscle regeneration and require clinical intervention to restore muscle function and prevent long-term deficits.Source: §I · p. 1
  • Design-of-Experiments parameter tuning is time- and material-intensive, while model-based control can require precise material characterization and simplifying assumptions.These constraints are challenging for biomaterials with non-Newtonian or time-varying properties.Source: §I · p. 1
  • ML-driven approaches can require large, structured datasets, generalize poorly across inks, and lack direct quantitative bead-thickness measurements.Some autonomous thickness-measurement methods also rely on user input, limiting near-real-time operation.Source: §I · p. 2
  • The proposed physics-based closed-loop framework uses 3D vision to measure deposited filament thickness and adapt printing pressure and needle height near real time.It is presented as not relying on pretrained models, large datasets, or material-specific formulations.Source: §I · p. 2

II. METHODOLOGY

The framework uses vision-based feedback to adapt DIW printing parameters toward a desired bead thickness. It directly controls pressure and adjusts needle height indirectly through a reduced-order physics-based model.

II. METHODOLOGY

  • The control loop measures bead thickness and adjusts printing parameters according to the difference between measured and desired thickness.The paper describes this feedback framework as an alternative to typical DoE workflows and material-specific fluid-dynamics modeling.Source: A. Proposed Framework and Problem… · p. 2
  • Printing pressure is controlled directly, while needle height is modeled as a function of pressure; other printing parameters remain constant.The needle-height relationship uses a reduced-order physics-based model.Source: A. Proposed Framework and Problem… · p. 2
  • The framework can begin at an arbitrary or predetermined pressure and adapt it based on measured bead thickness during printing.Adaptation is intended to mitigate effects from material changes and system dynamics.Source: A. Proposed Framework and Problem… · p. 3

B. Direct Control of Printing Pressure

The framework adjusts pressure using integral feedback on thickness error and uses a reduced-order physics-based model to adjust needle height as flow changes. Its vision pipeline calculates bead thickness along the needle path.

B. Direct Control of Printing Pressure

  • Integral control adjusts printing pressure from accumulated thickness-tracking error, with discrete updates summing error over sampling intervals.The reported experiments indicate sufficient accuracy and elimination of steady-state error.Source: B. Direct Control of Printing Pressure · p. 3

C. Indirect Control of Needle Height

  • Changing material flow affects bead thickness and deposited height, so needle height must be adjusted to prevent the needle from penetrating the bead.Source: C. Indirect Control of Needle Height · p. 3
  • Needle height is treated as dependent on deposition rate and constant transverse velocity, with pressure-driven flow modeled using Carreau rheology for non-Newtonian materials.The Carreau model is presented as more accurate for non-Newtonian behavior than simpler models such as the Power Law.Source: C. Indirect Control of Needle Height · p. 3
  • The flow estimate is used with conservation of mass to estimate deposited material height and set the needle height.The flow calculation uses a Carreau-fluid pipe-flow model and a numerically solved wall shear rate.Source: C. Indirect Control of Needle Height · p. 3
  • The vision pipeline scans and masks the print site, detects and joins bead contours, then calculates thickness at each point along the needle path.Source: Figure 3 · p. 4
  • The Dynamic Height Adjusting function approximates H=f1(Q(P)) as the linear relation H(P)=h0+αP, with parameters fitted to experimental data.The linear approximation is chosen to avoid a complex formulation considered unsuitable for real-time use.Source: C. Indirect Control of Needle Height · p. 4

D. Robotic Bioprinting Setup

The experimental system combines a robotic manipulator, pneumatic bioprinter, and structured-light 3D camera. ROS2 provides the control and communication environment for the hardware and sensing modules.

D. Robotic Bioprinting Setup

  • The setup uses a seven-degree-of-freedom KUKA robotic manipulator, a custom air-pressure-driven bioprinter, and a structured-light camera capturing RGB images and 3D point clouds.The camera's specified point-cloud resolution is 0.2 mm at a 300 mm working distance.Source: D. Robotic Bioprinting Setup · p. 4
  • ROS2-Humble connects the robot, camera, dispenser, pneumatic valve, and sensing and control modules; KUKA's Fast Robot Interface transmits commands at 100 Hz.Source: D. Robotic Bioprinting Setup · p. 4

E. Hardware Calibration and Registrations

Hardware calibration establishes the needle tip’s location and maps camera point-cloud data into the robot’s coordinate frame.

E. Hardware Calibration and Registrations

  • Pivot calibration locates the printer tip in the robot workspace, while eye-in-hand calibration transforms camera point-cloud data into the robot frame.Source: E. Hardware Calibration and… · p. 4

1) Pivot calibration:

Pivot calibration estimates the offset between the robot end effector and dispensing needle tip by rotating the printer around a stationary needle tip.

1) Pivot calibration:

  • The calibration records end-effector poses while rotating the printer around the stationary needle tip, then uses least squares to estimate the tip offset and pivot point.The offset relates each recorded end-effector position and rotation to the stationary pivot location.Source: 1) Pivot calibration · p. 4

2) Eye-in-hand calibration:

Eye-in-hand calibration estimates the camera-to-end-effector transform from robot and camera observations across multiple poses, enabling camera scans to be mapped into the robot world frame.

2) Eye-in-hand calibration:

  • The calibration estimates the missing spatial transform between the camera frame and robot end-effector frame.Source: 2) Eye-in-hand calibration · p. 4
  • At multiple robot poses, the camera measures the calibration board relative to itself while forward kinematics provides the world-to-end-effector transform.Source: 2) Eye-in-hand calibration · p. 4
  • Least squares on pose-pair equations estimates the transform, allowing camera scans to be mapped into the robot world frame.Source: 2) Eye-in-hand calibration · p. 5

F. Delineation and Thickness Calculation Algorithms

The autonomous vision method measures bead thickness intermittently for near-real-time closed-loop control, masking the printer tip to support online measurement.

F. Delineation and Thickness Calculation Algorithms

  • The vision-based method calculates bead thickness quickly and autonomously, making it suitable for near-real-time measurement and closed-loop control.It is intended to address limitations of offline scans, including slow processing, user inputs, and nozzle masking.Source: F. Delineation and Thickness… · p. 5

1) Step I - Scanning:

  • A camera scan captures the print site with the printing needle tip in view for masking in later processing.Source: 1) Step I - Scanning · p. 5

2) Step II - Grayscale conversion and mask definition:

4) Step IV - Edge detection:

The vision pipeline refines detected bead contours, overlays the printing path, and measures bead thickness along that path.

4) Step IV - Edge detection:

  • Canny edge detection identifies image contours where intensity gradients exceed a threshold, but its contours are discontinuous.The discontinuities require further refinement.Source: 4) Step IV - Edge detection · p. 5

5) Step V - Contour Tracing:

6) Step VI - Thickness calculation:

III. RESULTS & DISCUSSION

Across four configurations, the controller adjusted pressure and needle height to track desired bead thickness from arbitrary initial pressure values. Closed-loop trials converged, while open-loop trials showed larger errors.

  • Average convergence times ranged from 3.6 to 6.5 seconds, with convergence defined as thickness error staying below 0.5 mm.The four configurations used two target thicknesses and two initial pressures, with three repetitions each.Source: §III · p. 7
  • Three fitted fluid models matched rheology data but predicted different ideal needle heights, limiting fluid-dynamics modeling as a stand-alone control basis.The tested ultrasound gel showed shear-thinning behavior, with viscosity dropping at higher shear rates,.Source: §III · pp. 5, 7
  • For configurations 1–2, pressure increased to reduce under-deposition; for configurations 3–4, it decreased to reduce over-deposition.The plots show measured and desired thickness, controlled pressure and needle height, and thickness error over time.Source: Figure 5 · p. 6, Figure 6 · p. 6, §III · p. 7
  • Image processing took 0.38–0.45 seconds on average, below the camera’s 1.23–1.68 seconds, making capture time the larger timing component.The authors identify newer camera hardware as a possible route to further improvement.Source: §III · p. 7
  • All four configurations converged below 0.5 mm error, while open-loop errors reached 5 mm; converged-pressure standard deviation peaked at 0.05 bar.Figure 7 aggregates the three trials per configuration, showing mean and standard deviation; Table II reports configuration results.Source: Figure 7 · p. 6, §III · p. 7
  • A supplementary step-disturbance demonstration showed pressure and needle-height control tracking a changed thickness target during printing.Source: §III · p. 7

IV. CONCLUSION AND FUTURE WORK

The paper presents a physics-based, vision-feedback closed-loop framework for adapting DIW printing parameters to achieve desired bead thickness. Across four experimental configurations, it converged in 5.2 seconds on average; faster camera capture and VML-specific biomaterials remain future-work directions.

  • The framework replaces time- and material-demanding design-of-experiments parameter selection with closed-loop vision feedback and autonomous construct evaluation.Source: §IV · pp. 7–8
  • The four configurations converged to desired thicknesses in 5.2 seconds on average; camera capture averaged 1.36 seconds versus 0.43 seconds for processing.Table II is identified as reporting results for the four configurations.Source: §IV · p. 8, Table II · p. 8
  • Future work will test biomaterials designed specifically for VML treatment and improve performance through hardware improvements.Source: §IV · p. 8
Loading 2609.12159v1…