GenTact Toolbox : A Computational Design Pipeline to Procedurally Generate Context-Driven 3D Printed Whole-Body Artificial Skins

Carson Kohlbrenner, Caleb Escobedo, S. Sandra Bae, Alexander Dickhans, Alessandro Roncone · 2024

Read GenTact Toolbox : A Computational Design Pipeline to Procedurally Generate Context-Driven 3D Printed Whole-Body Artificial Skins by Carson Kohlbrenner, Caleb Escobedo, S. Sandra Bae, Alexander Dickhans, Alessandro Roncone (2024) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2412-00711.txt.

Introduction

Carson Kohlbrenner and co-authors built in 2024 a design pipeline that grows touch-sensitive skins to fit a specific robot and its job. Off-the-shelf skins are modular but ignore the machine's shape and setting. Their toolbox generates the skin's mesh from the robot's own geometry, then uses task simulation to place sensors where they earn their keep.

Multi-material 3D printing then fabricates the skin for almost any form. The team made six capacitive skins and fitted them to a research robot arm for a human-robot interaction test. The skins worked in place. The authors frame the shift plainly: from one-size-fits-all sensors to context-driven designs. They argue whole-body touch should be tailored like clothing, not bolted on like armor.

Catalog record

CREATORCarson Kohlbrenner, Caleb Escobedo, S. Sandra Bae, Alexander Dickhans, Alessandro Roncone
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-00711

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

GenTact Toolbox : A Computational Design Pipeline to Procedurally Generate Context-Driven 3D Printed Whole-Body Artificial Skins Carson Kohlbrenner, Caleb Escobedo, S. Sandra Bae, Alexander Dickhans, Alessandro Roncone Abstract Developing whole-body tactile skins for robots remains a challenging task, as existing solutions often prioritize modular, one-size-fits-all designs, which, while versatile, fail to account for the robot’s specific shape and the unique demands of its operational context. In this work, we introduce GenTact Toolbox , a computational pipeline for creating versatile whole-body tactile skins tailored to both robot shape and application domain. Our method includes procedural mesh generation for conforming to a robot’s topology, task-driven simulation to refine sensor distribution, and multi-material 3D printing for shape-agnostic fabrication. We validate our approach by creating and deploying six capacitive sensing skins on a Franka Research 3 robot arm in a human-robot interaction scenario. This work represents a shift from “one-size-fits-all” tactile sensors toward context-driven, highly adaptable designs that can be customized for a wide range of robotic systems and applications. The project website is available at https://hiro-group.ronc.one/gentacttoolbox I Introduction Whole-body tactile feedback assists humans in effectively interacting with the environment, developing social connections, and avoiding danger [ 1 , 2 ] . Similarly, robots equipped with whole-body tactile sensors benefit from touch feedback in exploration, interaction, and task execution in unstructured environments [ 3 , 4 , 5 ] . To this end, there has been a significant body of work in developing whole-body tactile skins, i.e. arrays of sensors designed to give a robot the sense of touch over the full surface of its body [ 6 , 7 , 8 ] . However, state-of-the-art methods are one-off solutions with limited adaptability, and, to the best of our knowledge, no platform currently exists that enables roboticists to design or customize purpose-built, whole-body skins tailored to their specific robot or use case. Interestingly, a robot’s operational context and shape can directly influence critical design parameters of a whole-body artificial skin, such as sensing resolution, surface coverage, the skin’s softness/compliance, or the bandwidth at which data needs to be processed to successfully complete the task. For example, precise manipulation in clutter (e.g. [ 9 , 10 , 3 , 11 ] ) requires high-resolution data to localize force as opposed to a binary and low resolution detection of contact needed for safe physical human-robot interaction (pHRI) (e.g. [ 12 , 13 , 14 , 15 ] ). Fig. 1 : The computational pipeline presented in this work, GenTact Toolbox , generates form-fitting and adaptable whole-body tactile sensors. GenTact Toolbox uses the 3D model of a given robot (a) and a user-generated heat map (b) to create digital meshes of sensor arrays (b) that can be 3D printed as functional tactile sensors (c). Fig. 2 : The GenTact pipeline for designing form-fitting and adaptable tactile skins is composed of three stages: procedural generation, simulation, and fabrication. The procedural generation stage (left) generates an initial distribution of sensors that are then passed into the simulation stage (bottom right) to be evaluated and improved based on the task they are used for. Finally, the sensors are connected internally in the fabrication stage (top right) to be printed and deployed on the real robot. Another gap this work seeks to address is that current state of the art systems have a high barrier to entry for end users. The literature focused on designing sensors that conform to robot geometries with modular or flexible designs that wrap over a robot surface [ 8 , 16 , 17 , 18 , 19 , 20 ] . While wrapping sensors to conform to robot geometry affords generalization after fabrication, it can be problematic at the hardware and software integration levels because it requires manual assembly and localization of each sensor [ 21 ] . Of note, robots come with a 3D model that contains precise link dimensions and geometry that can alleviate these integration issues if applied across both the pre-fabrication and post-fabrication stages. It is therefore possible to leverage this 3D model to inform the design of a new class of tactile sensors that are generated for a specific embodiment and towards a specific application . This process, inspired by procedural generation techniques used in computer graphics and video games, can be further combined with generalized fabrication processes such as 3D printing. This in turn allows us to bring the flexibility of procedural generation from the design stage to the manufacturing itself, and enables customizable, form-fitting, functional, and rapidly deployable tactile skins. In this work, we present GenTact Toolbox , a computational design pipeline that produces form-fitting and context-driven tactile sensors tailored to the unique demands of individual robots and application domains ( Fig. 1 ). GenTact Toolbox has three main components: a procedural mesh generation stage that algorithmically designs form-fitting and functional tactile sensing skins, a task-driven simulation stage that optimizes sensor placement for a given application, and a multi-material 3D printing phase to fabricate generated skins as capacitive sensors. The first stage of our pipeline has designers indicate the regions of skin coverage and sensor placement over a robot’s 3D model with virtual heat maps. The designer can then vary the skin thickness, sensor size, and sensor density within these heat maps by adjusting configurable scalar parameters. In the second stage, the generated skin sensors are simulated in Isaac Sim to generate contact data for a given user-defined task. We use a customizable heuristic to feed this simulation data back into the first stage to create an optimized heat map with high density in contact-rich regions. Lastly, the optimized sensor layout is used to produce a 3D printable tactile skin with individually addressable contact points. We demonstrate the efficacy of our approach by generating six unique sensing units to cover a Franka Research 3 (FR3) robot arm and testing them in the real world for a pHRI task. We also generate full-surface skin designs for the Unitree humanoid H1 and quadruped Go2 to highlight the breadth of our approach. Our experimental results reveal that procedural generation is a capable tool for the design of future whole-body tactile skins. The primary contributions of this paper are: (i) a procedural generation approach to produce tactile sensor designs; (ii) an open-source context-driven design pipeline to procedurally generate and optimize 3D printed capacitive tactile sensors; (iii) an evaluation of our pipeline for instantiating sensors for a pHRI use case scenario. II Related Work II-A Design Challenges in Whole-Body Tactile Sensing Previous implementations of whole-body skin sensors have focused on resolving the geometric constraints of covering robots during post-fabrication . These sensors are often rigid modular sensors [ 17 , 16 , 22 , 23 , 21 ] or flexible electronics [ 18 , 19 , 24 ] that are designed to work in tandem with additional units. These modular sensors can wrap around the surfaces of robots to cover simple and partial curvatures while also being highly reproducible. However, they require a full re-design when the applied geometry surface changes (e.g., fitting over smaller, more complex surfaces) or when needing to alter sensor density to place the robot in a different work context. Flexible sensors can offer a tighter fit by omitting rigid

Read the complete text