Designing a soft robot that replicates jellyfish locomotion
To imitate an animal's motion it helps to measure it first. This project began with a video of a swimming jellyfish: a program extracts the bell contour in each frame and obtains the profile, the extreme positions and the speeds of the contraction. With those data, a pneumatic silicone tentacle was designed whose initial shape matches the animal's profile at its most open position, and a first prototype with four tentacles and a membrane was built. The prototype reproduces the positions, but its motion is slower than the animal's.
Context
A jellyfish propels itself by contracting its bell to expel water and relaxing it to refill. Although the bell is a continuous membrane, its muscle fibers are organized in radial bands. That observation suggests a simple architecture for a robot: several bending actuators arranged radially and joined by a passive membrane.
Measuring the motion
The video was processed frame by frame. The background is removed, edges are detected, and a line-by-line sweep inside an analysis window extracts the contour of half the bell.
Detected contour and analysis window on one video frame
Assuming rotational symmetry, the contour yields the internal volume of the bell and the area of its mouth at each instant. The derivative of the volume gives the expelled flow rate, from which thrust is estimated.
Tentacle design
The tentacle is a bending actuator of the kind used in soft robotics: a row of chambers over an inextensible layer. Its rest geometry was fitted to the profile measured in the video.
Initial actuator profile. In red, the jellyfish profile obtained from the video
Before fabrication, the direction of deformation was checked with a voxel simulation.
Tentacle bending sequence simulated in VoxCAD
Final tentacle geometry and chamber detail
Actuation
Two three-way valves, one inlet and one outlet, feed the four tentacles in parallel. Opening time sets the position, and reservoir pressure sets the speed.
Pneumatic circuit driving the four tentacles
Prototype
The tentacles were cast in Ecoflex in two halves that were then glued together. They were mounted on a printed head that houses the valves, and the assembly was covered with a membrane.
Pouring the silicone and the two cured halves of the tentacle
Tentacles joined to the head and the actuation system
Results
Profile of the jellyfish and of two tentacles at the most open and most closed positions
At the extreme positions, the tentacle profile approximates that of the jellyfish. In time, it does not. Closing is about 15% slower, and the return takes between 3 and 4 s against about 2 s in the animal. The return depends only on the elasticity of the silicone and the paper layer, and on how fast the air leaves.
The thrust analysis of the real jellyfish, done from the video, was inconclusive. Flow rate is obtained by differentiating a noisy signal, and only the propulsion peak at closure can be distinguished.
Volume, mouth area, flow rate and propulsion estimated from the video
What is missing
The prototype was tested in air, not in water. The tentacles, designed curved, sag under their own weight; making them straighter is proposed. The valves that fit inside the robot do not withstand the pressure needed for fast inflation, so placing them outside is suggested. Manufacturing was not very repeatable and took up a good part of the project's time. An analytical relation between pressure and tentacle bending is also missing.
How it fits in Robiolab
This project shows the method the lab associates with bioinspiration: measure the organism, extract the principle and design with it, rather than copy the appearance. The four proposed improvements were the starting point of the swimming robot implementation the following year, which took up these recommendations.