This video details the ongoing development of "Tatbot," an innovative robot designed for tattooing. The creator showcases the system's core components, including Trossen robotic arms, NVIDIA Jetson edge computing, and a sophisticated multi-camera setup (RGB and RealSense depth cameras). A primary focus is on solving the complex computer vision challenge of real-time 3D skin reconstruction and projecting 2D tattoo designs onto these deformable 3D surfaces using deep learning models like Master and VGGT, along with NVIDIA Warp for design warping. The creator also demonstrates inverse kinematics solutions using Pyroki for precise robot control. Unlike existing open-loop tattoo machines, Tatbot aims for 6D control to enable tattooing on curved body parts. The video highlights a philosophy of self-funding, open-source development, and targeting high-value automation tasks, with plans for a dual-arm system where one arm tattoos and the other performs wiping actions using advanced Vision-Language-Action (VLA) models.
Exploring the Tatbot: Revolutionizing Tattooing with Advanced Robotics
0:45 Project Introduction - Tatbot: The creator announces "Tatbot," a new project focused on developing a robot capable of performing tattoos, emphasizing its innovative capabilities.
0:50 Hardware Overview: The robot is constructed with 80/20 aluminum framing, two Trossen Widow XAI arms (offering 1.5kg payload and 1mm repeatability), and an NVIDIA Jetson AGX Orin for edge computing.
0:55 Integrated Camera System: The system features multiple Power over Ethernet (PoE) IP cameras and two Intel RealSense D405 cameras for RGB and depth data, aiming for robust real-time environmental awareness.
1:20 Off-the-Shelf Tattooing Tools: Tatbot is designed to use standard, off-the-shelf tattoo wands/machines and pre-made needle cartridges, allowing for flexibility and avoiding custom end-effector designs.
1:40 Critique of Existing Tattoo Robots: The creator reviews current tattoo robots, such as 3D printer-based rigs (1:40) and the Black Dot company's system (1:50). These systems are noted for their open-loop, XYZ-only control, limitation to flat surfaces, and lack of force feedback.
2:00 Tatbot's 6D Control Objective: A key innovation of Tatbot is its ability to perform 6D control (XYZ position plus yaw, pitch, and roll orientation), enabling tattooing on complex, curved body surfaces.
2:10 Real-time Skin Reconstruction: To achieve 6D control on deformable skin, the robot uses its multi-camera setup for real-time 3D reconstruction. The creator discusses deep learning models like "Master" and "VGGT" for this purpose (2:25), demonstrating their performance (3:00, 3:15).
3:30 2D to 3D Design Projection: NVIDIA Warp's cloth simulation is utilized to project flat, AI-generated tattoo designs onto the reconstructed 3D skin mesh, translating the 2D art into a 3D path for the robot.
3:45 Inverse Kinematics (IK) for Precision: Pyroki, a JAX-based library, is employed to solve the inverse kinematics problem, calculating the precise joint angles required for the robot's end-effector (tattoo pen) to follow the intricate 3D design path (3:50).
1:01:50 Business Philosophy & Independence: The creator aims to self-fund the project, highlighting a preference for high-value manipulation tasks like tattooing over lower-value automation, and seeking to avoid venture capitalist influence to maintain creative freedom.
1:08:49 Robotics Foundation Models for Auxiliary Tasks: While precise tattooing relies on IK, less precise tasks like "wiping" (cleaning excess ink) will leverage Vision-Language-Action (VLA) models like NVIDIA Groot or Physical Intelligence's PI-0. This hints at a dual-arm setup (1:15:09).
1:27:06 Open-Source Development: All code for Tatbot will be open-source, promoting collaboration and transparency rather than proprietary secrecy.
1:28:03 First Robot Tattoo & Early Release: The creator showcases an early, imperfect robot-tattooed "circle" using a Sharpie, embracing the philosophy of releasing early and often, even if "embarrassing."