Active entanglement enables stochastic, topological grasping

Abstract

Grasping, in both biological and engineered mechanisms, can be highly sensitive to the gripper and object morphology, as well as perception and motion planning. Here, we circumvent the need for feedback or precise planning by using an array of fluidically actuated slender hollow elastomeric filaments to actively entangle with objects that vary in geometric and topological complexity. The resulting stochastic interactions enable a unique soft and conformable grasping strategy across a range of target objects that vary in size, weight, and shape. We experimentally evaluate the grasping performance of our strategy and use a computational framework for the collective mechanics of flexible filaments in contact with complex objects to explain our findings. Overall, our study highlights how active collective entanglement of a filament array via an uncontrolled, spatially distributed scheme provides options for soft, adaptable grasping.

Keywords: entanglement; filaments; soft actuators; soft robotic grasping; soft robots.

Fig. 1.

Fluidically actuated entangling filaments. (A) Photograph of an entangling filament gripper consisting of 12 hollow elastomeric filaments in a resting state and pneumatically actuated around a house plant. (B) Schematic of filaments at ambient and increased internal pressure. (C) Schematic of the entanglement of two nearby filaments. The filaments are not entangled in their rest state or at low pressures. At low pressures, the filament begins to curl in a plane, and as the internal pressure approaches the operating pressure (in this case 172 kPa), the filaments bend out of their initial plane and start to entangle with nearby filaments. (D) Physical and simulated entanglement examples with contacts between filaments and the object (eight-branch tree) indicated. Contacts are color-coded and grouped by individual filaments. (E) Photographs of an array of 12 filaments activated by an internal pressure of 172 kPa and entangled around neighboring filaments and various objects.

Fig. 2.

Spatial distribution of contacts and entanglement. (A) Schematic of four entangled filaments and a spherical bounding volume used to isolate and evaluate local metrics, such as the spatial density of filaments and the localized entanglement density, the results of which are presented in D and E. (B) The spherical bounding volume is projected onto a plane, and the number of crossings between filament center lines is used as an indicator of entanglement. The average over all projection directions is used to find an ACN. (C) Micro-CT–based three-dimensional (3D) reconstructions of the entangled filaments and objects used to extract the position and shape of each filament. Each filament is uniquely colored to distinguish individuals among the given 12-filament array. (D) Spatial density of filaments, calculated based on the number of filaments that occur within a spherical bounding volume with a 20-mm radius. The colors correspond to the number of filaments inside the local bounding volume. (E) Localized ACN of the filaments, calculated as an average number of filament crossings over all projections of a spherical bounding volume with a 20-mm radius. The colors correspond to the ACN within the local bounding volume. (F) A 3D rendering from micro-CT scans of entangled filaments, with filament–filament contacts highlighted in blue and filament–object contacts highlighted in red. The entanglement examples and objects are the same as those shown in Fig. 1E and panels C–E of this figure. (G) Plots of the probability density of the entanglement number (red) and number of contacts (blue) in a 20-mm-radius spherical bounding volume at each point of the array from the scanned grasps shown in C–F. Additional plots for spatial distribution and area of contact are included in SI Appendix, Fig. S7–S9.

Fig. 3.

Collective entanglement as a stochastic grasping strategy. (A) Photographs showing a sequence of entanglement grasping tests conducted with a UR5 robotic arm. (B) Success rate of grasping tests performed with various objects that were centered directly below an array of 12 filaments. Objects include the eight-branch tree shown in B, a 38-mm-diameter cylinder, a 10-cm-diameter sphere, and four objects from an adversarial object set (12). Additional object information is included in SI Appendix, SI Text. All simulated tests were performed with the eight-branch tree. (C) Success rate of a grasping test performed with various objects with increasing horizontal offsets between the vertical center line of the array of filaments and the target object. (D) Grasping-test success rates for a branched object with varying filament spacing and increasing horizontal offsets. (E) Grasping-test success rates for a branched object with varying spatial density (branch length) and increasing horizontal offsets. (F) Grasping-test success rates for a branched object with four different weights and increasing horizontal offsets. (G) Phase space of grasp success rate predicted by simulations of filaments entangling with the branched test object. Each plot represents a different object weight. Sweeps of object spatial density and filament spatial density are shown within each plot. The phase-space locations corresponding to data in D, E, and F are indicated on the plots.