Grasping by Spiraling: Reproducing Elephant Movements with Rigid-Soft Robot Synergy

TL;DR

This paper addresses replicating elephant trunk grasping with a hybrid rigid–soft robot by leveraging a logarithmic spiral geometry to realize bending and twisting primitives. The authors integrate a rigid Franka arm with a cable-driven soft manipulator and develop a length-based forward kinematic model anchored to the spiral form $r = a e^{b \theta}$, enabling controlled shape evolution from tip to base. They demonstrate nine of seventeen documented elephant grasping strategies and validate the model with motion-capture experiments that yield RMSEs in the range of a few to a few tens of millimeters, highlighting dynamic curvature propagation and stiffness modulation as core capabilities. The work advances cross-scale, adaptable grasping with reduced control complexity and lays the groundwork for future force-controlled, learning-based extensions in unstructured environments.

Abstract

The logarithmic spiral is observed as a common pattern in several living beings across kingdoms and species. Some examples include fern shoots, prehensile tails, and soft limbs like octopus arms and elephant trunks. In the latter cases, spiraling is also used for grasping. Motivated by how this strategy simplifies behavior into kinematic primitives and combines them to develop smart grasping movements, this work focuses on the elephant trunk, which is more deeply investigated in the literature. We present a soft arm combined with a rigid robotic system to replicate elephant grasping capabilities based on the combination of a soft trunk with a solid body. In our system, the rigid arm ensures positioning and orientation, mimicking the role of the elephant's head, while the soft manipulator reproduces trunk motion primitives of bending and twisting under proper actuation patterns. This synergy replicates 9 distinct elephant grasping strategies reported in the literature, accommodating objects of varying shapes and sizes. The synergistic interaction between the rigid and soft components of the system minimizes the control complexity while maintaining a high degree of adaptability.

Figures (5)

• Figure 1: The schematic graph demonstrates our hybrid system's capability to mimic the elephant motion primitives. (a).An overview of the proposed hybrid system compared to typical elephant motions. The system comprises a Franka Emika Panda robotic arm and a cable-driven soft continuum manipulator with a trunk-tip gripper. Three typical configurations are present: bending when lowering the head (middle), twisting with the head up (right), and resting under gravity (left). (b). The exploded CAD view of the motor housing demonstrates the detailed placement of the actuators. (c). The rotation of the steering gear creates a two-way open and close for the nose-tip gripper. (d),(e) The schematic graph of the point-to-point elephant grasping procedure, including reaching, prehension, transport, and release phases. (The real trunk demonstration images are adapted from the video in dagenais2021elephants). During the reaching phase, the primary deformation is elongation, which facilitates an extended reach. During prehension, small objects are grasped through an open-and-close motion of the trunk tip, while larger objects exceeding the capacity of the nose tip are wrapped using the distal portion of the trunk, achieved through bending and/or twisting, depending on the object's position. In the transport phase, a curvature propagation would occur towards the proximal end to secure the object by inward bending. Finally, during the releasing phase, an opposite curvature propagation assists in releasing the object at the target location.
• Figure 2: (a). Three sprials at different scales formed by varying the scale factor $a$. (b). By varying the growth parameter $b$, we can change the growth rate of the spiral. (c). The compactness parameter $k$. (d). The discrete parameter $\Delta\theta$. (e). The optimal parameters are selected for the geometry constraint of our soft manipulator and its end-effector gripper. (f). The final dimensions of our soft manipulator are determined based on the selected optimal parameters. (g),(h): The qualitative comparison of bending and twisting between the analytical model (above) and prototype (below) under the same actuation amount in different gravity directions.
• Figure 3: (a). Schematic graph of the motion capture validation experiment for the model. (b). The visualization of the motion capture results in the vertical plane for different bending shapes, and the quantitative result shows the averaged error along shapes (c) and sections (d). (e). The visualization of the motion capture results in the resting shapes under different gravity angles, and the quantitative result shows the averaged error along shapes (f) and sections (g).
• Figure 4: Demonstration of the hybrid system's capability of reproducing the 9 distinct elephant grasping strategies. The listed strategy and elephant demonstration are derived from studies in dagenais2021elephants. Key shapes with different actuation patterns are determined from the study and pre-defined by the forward kinematic model. The snapshots of our prototype demonstration showcases the ability of a successful prehension and transportation of geometric objects. Full video demonstration can be seen in the supplementary files.
• Figure 5: (a). Schematic graph of the spiral with optimal parameters. (b). A two-section arm hangs freely with an incline angle $\theta_0$ under gravity. Notably, only the upper half of the body is shown here and in (c) for simplification. (c). The generalization of force equilibrium to multiple sections. (d). Bending deformation. The points $A, B, C, D, E, G$ represent the positions of the holes where the cables run through, while the colored lines represent cables. $Axis 1$ is obtained by transforming the vector $EG$ to pass through point $E$. When tightening cable $ABCD$, the section on the right would rotate around $Axis 1$. When a series of sections rotate around their own axis (which is parallel with each other), a bending effect is induced where all the joint rotation angles are in the same plane. (e). Twisting deformation. On top of the bending effect, if the second cable $FG$ is subsequently tightened, in order to maintain the length of $ABCD$, the section on the right would rotate along $Axis 2$, which is defined by the vector $BE$. When a series of sections rotate around their own axis (which is not parallel with each other), a twist is induced where the joint rotation angles would not be in the same plane.