Parametric design of grippers that handle corals without breaking them
Reef restoration requires fragmenting, moving and planting thousands of coral pieces, and today it is done by hand. Automating that task calls for a gripper that holds a branching, irregular, fragile object without fracturing it. This work approaches the problem through the shape of the gripper. An automated pipeline generates hundreds of geometries, simulates them against a scanned coral fragment, and selects those that hold with enough force while keeping contact pressure low. Surfaces with a wavy texture reduced peak pressure by up to 40% compared with flat ones.
Context
Acropora cervicornis, staghorn coral, is one of the central species in Caribbean reef restoration. Its skeleton is porous aragonite, and the compressive strength reported in the literature ranges from 12 to 81 MPa depending on porosity and growth direction. For corals weakened by bleaching, the thesis adopts a failure threshold close to 19 MPa.
What breaks the coral is not the total gripping force but the local pressure where the gripper touches a protrusion. The design criterion is therefore to spread the load.
A plaster coral
Testing with live corals is not viable for ethical and logistical reasons. An analog was built: a 3D-scanned fragment was printed in PLA, a silicone mold was made from it, and type III dental stone was cast in the mold.
Casting and demolding the coral fragment in plaster
Based on the literature, the wet plaster has a strength of about 20.7 MPa, very close to that of the bleached coral skeleton. A gripper that does not fracture the plaster specimen therefore has a margin against the coral.
Automatic generation and simulation
The gripper is defined by a few parameters: length, curvature, and the terms of a sinusoidal texture on the contact surface, with its amplitude and frequency.
Gripper geometry generated from parameters
Tree of parameter combinations defining the design space
Grasshopper generates each geometry, and a Python program sends it to ANSYS Mechanical, sets up contacts and boundary conditions, runs the simulation and collects the results. Two routines remove the manual work: one recognizes the gripper's mounting holes and the other infers by proximity which surfaces come into contact. They worked on all 324 variants of the main set.
To make the sweep feasible, the coral mesh was simplified from 51,750 to 1,688 faces.
Progressive simplification of the coral fragment mesh
Simplification cut computation time by an order of magnitude and changed reaction forces and overall pressure by less than 2%.
Contact pressure on the fragment, with the original mesh and with the simplified mesh
Results
In total 448 simulations were run, first against a cylinder and then against the coral fragment.
Contact pressure against total gripping force for the full set on the coral. Below, a close-up of the region of interest
- The sinusoidal texture is the most influential factor. It increases effective contact area and reduces peak pressure by up to 40% relative to a flat surface.
- There is a family of safe designs that holds with more than 5 N, the force estimated for stability against currents, without exceeding the failure threshold.
- The best design on the coral applied 51.6 N with a contact pressure of 3.1 MPa.
Best-performing design for the coral fragment, printed in PETG and PLA
Physical validation
The best designs were printed and mounted on a bench with a linear actuator. The contact surfaces were covered with pressure-sensitive film, which leaves a color mark proportional to the pressure received.
Gripper on the test bench with pressure film, open and closed
Pressure-sensitive film after the test with the plaster coral model
The regions marked on the film coincide with the pressure concentration regions predicted by the simulation.
What is missing
The comparison between film and simulation concerns spatial distribution. The film is a semi-quantitative instrument and the work does not report measured pressure values against simulated ones. No tests were run with real coral or underwater, and the argument for equivalence between plaster and coral rests largely on literature data. Damage to living tissue, which is more fragile than the skeleton, was not assessed either. A single fragment was studied, and morphological variability between colonies is large.
How it fits in Robiolab
This project uses generative design, that is, automatically exploring a space of shapes and letting simulation select, which is also the method by which evolution produces the morphologies that bioinspired robotics imitates. The object being handled is, moreover, a living organism. The infrastructure it leaves, a pipeline that connects parametric geometry to finite element analysis without manual intervention, can be reused for other designs in the lab where contact is the central problem, such as wearable-device interfaces and soft actuators. It has a direct precedent in the generative design interface for soft actuators.