A Jellyfish Cyborg: Exploiting Natural Embodied Intelligence as Soft Robots

TL;DR

This work developed an integrated muscle electrostimulation and 3D motion capture system to quantify both spontaneous and stimulus-induced behaviors in Aurelia coerulea jellyfish and successfully predicted future movements based on the current body shape and natural dynamic patterns of the jellyfish.

Abstract

Jellyfish cyborgs present a promising avenue for soft robotic systems, leveraging the natural energy-efficiency and adaptability of biological systems. Here we demonstrate a novel approach to predicting and controlling jellyfish locomotion by harnessing the natural embodied intelligence of these animals. We developed an integrated muscle electrostimulation and 3D motion capture system to quantify both spontaneous and stimulus-induced behaviors in Aurelia coerulea jellyfish. Using Reservoir Computing, a machine learning framework, we successfully predicted future movements based on the current body shape and natural dynamic patterns of the jellyfish. Our key findings include the first investigation of self-organized criticality in jellyfish swimming motions and the identification of optimal stimulus periods (1.5 and 2.0 seconds) for eliciting coherent and predictable swimming behaviors. These results suggest that the jellyfish body motion, combined with targeted electrostimulation, can serve as a computational resource for predictive control. Our findings pave the way for developing jellyfish cyborgs capable of autonomous navigation and environmental exploration, with potential applications in ocean monitoring and pollution management.

Figures (10)

• Figure 1: Overview of Jellyfish Cyborg System: ( A) Electrode-mounted jellyfish with a floating tethered system specially designed for this study. ( B) Pulse width modulation (PWM) signals used for stimulating the jellyfish muscles, which replicate the neural commands. ( C) Schematic diagram of the proposed floating tethered system. ( D) Electrodes equipped with LED stimulus indicators and wires. ( E) Custom-built electrical stimulator, utilizing a Raspberry Pi Pico W. The parameters of the PWM signals are adjusted by a Python program that integrated into the stimulator.
• Figure 2: Experimental Animals and Setup: ( A) Aurelia coerulea medusae. ( B) Visible Implant Elastomer Tag under UV light. R*, Y*, O*, and B* in the figure correspond to the colors of the markers, with 1 and 2 indicating the positions of the outer and inner markers, respectively. ( C) Measurement system with a high-speed camera. ( D) One camera captured the top view, while two mirrors were positioned on the sides of the tank to enable synchronized, simultaneous recording from three different angles. ( E) Overlaid snapshots of camera images were taken every 6 seconds. The floating locomotion of jellyfish in a tank was reconstructed and quantified in three-dimensional (3D) space by estimating marker positions using DeepLabCut (DLC, nath2019using, Fig. S14). ( F) The 3D motion of the floating jellyfish was reconstructed. ( G) Time-series data of radial (top) and coronal (middle) lengths on the jellyfish body, along with 3D velocity in the defined BodyFrame (bottom), were calculated and presented (see Method for details).
• Figure 3: Spontaneous Pulsatile Floating Pattern: ( A) 3D spontaneous floating trajectories in a water tank ($N=6$ animals, 37 trials). Trajectories in the $x$-$z$ plane (top) and $x$-$y$ plane (bottom). Colors on the maps indicate the start and end of the measurements. ( B), ( C), and ( E) Distribution of power spectrum density (left column), duration (center column), and size (right column) for radial length (B), coronal length (C), and BodyFrame velocity (D). The vertical and horizontal axes are shown in log scale. The colors repsesent different lengths and velocities: R1-R2 (red), B1-B2 (blue), O1-O2 (orange), Y1-Y2 (green), R1-B1 (purple), B1-O1 (cyan), O1-Y1 (green), Y1-R1 (magenta). $x_{BF}$ velocity (crimson), $y_{BF}$ velocity (dark-green), $z_{BF}$ velocity (lime-green). Each dotted line represents a linear regression line in the corresponding region, and $\alpha$ is the value of the power exponent.
• Figure 4: A Representative Example of the Floating Pattern Induced by Electrostimulation $\tau=2.0$ s (supplementary video S3): ( A) Locomotion trajectory of a jellyfish (top view) during the 30-second electrostimulation period. Overlaid snapshots are shown at 2.0-second intervals of the stimulation cycle, with higher transparency close to the stimulation start time. ( B) Time series data of radial and coronal lengths, as well as BodyFrame velocities, including periods before, during, and after electrostimulation (yellow: before 15 s, white: stimulation 30 s, sky-blue: after 15 s). The gray area within the 30-second stimulus interval represents a duration of 0.1 seconds for a PWM signal input with a 2.0-second period. The pink highlights indicate one cycle (2.0 s) of electrostimulation in (C)-(E) below. ( C), ( D), ( E) Top row: Velocity changes in the $x$ (crimson), $y$ (dark-green), and $z$ (lime-green) directions in the BodyFrame during one stimulus cycle (2.0 s). Middle row: Jellyfish's position at the beginning and end of stimulation, as well as the time evolution of the velocity vector during electrostimulation in the $x$-$z$ plane (side view) of the tank. The velocity vectors change in shade of black from the beginning to the end of the 2.0 s stimulation. The velocity vectors in the $x$, $y$, and $z$ axes of the BodyFrame are also shown in each color. Bottom row: Identical data in the middle row, but in the $x$-$y$ plane (top view).
• Figure 5: Spatiotemporal Pulsatile Floating Pattern Analysis for Electrostimulation: $N=5$ subjects. (A) Time-series data in four stimulus input conditions: $\tau=0.5$, 1.0, 1.5, 2.0s, and control condition (without stimulation with electrodes and wires). For each condition, we conducted 5 trials, with the order of conditions randomized. To ensure consistency, we fixed the wires between R1 and Y1 and between O1 and B1. We only applied a PWM signal to the electrode between O1 and B1. The gray area within the 30-second stimulus interval represents a duration of 0.1 seconds for the PWM signal input, with each period $\tau$. (B) Frequency response analysis of the coronal lengths in four stimulus input (the red dotted lines show input frequencies) and control conditions. (C) We analyzed the power spectrum density (PSD distribution) for each spatially distributed radial and coronal length. (D) Detailed display of frequency analysis (O1-B1) for four stimulation input conditions, as well as cased without electrostimulation with electrodes and wires (w/o stim column) and spontaneous floating motion without electrodes and wires (free swim column), coresponding to Fig. \ref{['fig:spon_exp']}. Vertical and horizontal axes indicate input and output frequencies, respectively, and color shading indicates PSD. The red squares represent the input frequencies for the four conditions.