Speaker
Description
Introduction
Over the past two decades, robotic additive manufacturing (RAM) has been introduced into fields such as aerospace and construction, and multiple studies have shown the potential of RAM for Regenerative Medicine (RM). However, hardware limitations of current industrial robots reduce the print resolution of small volume RAM hindering the wider adoption into RM. Eye-in Hand (EH) technology, where the robotic arm is equipped with a camera at the end-effector, has been proposed to improve the resolution of robotic arms. Recent studies have used EH robot for in-situ bio-printing1,2, but the possibility to fabricate complex volumetric scaffolds using this technology has not been explored enough. Here, we developed an EH enabled RAM system to demonstrate the potential of the technology to fabricate small scale tissue engineering scaffolds with complex architectures overcoming the robotic hardware errors inherent to current RAM systems.
Methods
RAM
The RAM system used in this study is based on a 7 degree of freedom robot (Xarm7) coupled with a conventional thermoplastic filament extruder. The integration of the robot and extruder was handled by an updated version of RAVEN3, which is an open-source RAM package based on Robot Operating System (ROS2) and Moveit2. A newly developed real-time closed loop control system based on Moveit servo package was integrated into RAVEN for making use of data generated by EH to correct position errors of the robot.
EH
The EH system consists of a black and white camera (B0332, ArduCam) operating at 100 frames per second mounted close to the extruder pointed towards the tip of the nozzle. A computer vision is based on a pre-trained deep Convolutional Neural Network model (YOLOv8, Ultralytics Inc), which was tuned to identify printed filaments in the images captured by the camera. This system identifies the position of the filament in real-time which is then communicated to the RAVEN package, which controls the robot.
Scaffold design
The positional error of the robotic hardware was most pronounced when the orientation of the end-effector was changing. This prohibits printing of non-planar branched structures in 3D where RAM has to join multiple segments printed with different orientation of the extruder. Henceforth, simple designs with non-planar branching paths were used to test the efficacy of EH based position correction for improving printability of branched structures.
Results and Discussions
A novel open-source EH-enabled RAM and a proof of concept, closed-loops error correction system was developed for printing 3D branched structures. The printability improvement of branched structures offered by this approach demonstrates the need for developing complete real-time closed loop control architectures for small scale RAM. Further, the new tools offered by artificial intelligence technologies could make intelligent/adaptive RAM more accessible and efficient for RM applications.
References
1) Hu, J. et al. IEEE Robot Autom Lett (2024)
2) Jeong, S. H. et al. Adv Mater Technol 9, (2024).
3) Fucile, P., David, V. C. et al. Virtual Phys Prototyp 19, (2024).
96086716146