Authors: Zach J. Patterson, Emily Sologuren, Levi Cai, Daniel Kim, Alaa Maalouf, Pascal Spino, Daniela Rus
Organizations: Mechanical and Aerospace Engineering, Case Western Reserve University, Cleveland, 44106, USA. · Computer Science and Artificial Intelligence Laboratory, Massachusetts Institute of Technology, Cambridge, 02139, USA. · Applied Ocean Physics and Engineering Department, Woods Hole Oceanographic Institution, Woods Hole, 02543, USA. · Computer Science, University of Haifa, Haifa, 3498838, Israel.
Autonomous robots can transform how we observe marine ecosystems, but close-range operation in reefs and other cluttered habitats remains difficult. Vehicles must maneuver safely near animals and fragile structures while coping with currents, variable illumination and limited sensing. Previous approaches simplify these problems by leveraging soft materials and bioinspired swimming designs, but such platforms remain limited in terms of deployable autonomy. Here we present a sea turtle-inspired autonomous underwater robot that closed the gap between bioinspired locomotion and field-ready autonomy through a tightly integrated, vision-driven control stack. The robot combines robust depth-heading stabilization with obstacle avoidance and target-centric control, enabling it to track and interact with moving objects in complex terrain. We validate the robot in controlled pool experiments and in a live coral reef exhibit at the New England Aquarium, demonstrating stable operation and reliable tracking of fast-moving marine animals and human divers. To the best of our knowledge, this is the first integrated biomimetic robotic system, combining novel hardware, control, and field experiments, deployed to track and monitor real marine animals in their natural environment. During off-tether experiments, we demonstrate safe navigation around obstacles (91% success rate in the aquarium exhibit) and introduce a low-compute onboard tracking mode. Together, these results establish a practical route toward soft-rigid hybrid, bioinspired underwater robots capable of minimally disruptive exploration and close-range monitoring in sensitive ecosystems.
Figures & tables
Figure 1: Sea turtle–inspired robot for autonomous marine observation. a, A photograph of the turtle robot swimming in the New England Aquarium in Boston, MA. b, The turtle robot uses biomimetic soft-rigid flippers and visual feedback for tracking sea animals autonomously. c, Superimposed frames of robot swimming autonomously and untethered around coral in aquarium tank (the orange line denotes the path of the robot).
Figure 2: Autonomous animal following. a, Time series showing the robot’s point of view (POV) with target segmentation masks (green overlays) during autonomous sea turtle tracking over ∼ 12 seconds. b, Target centroid trajectories throughout a minute-long tracking sequence, with corner indices marking trajectory extremities (top left, top right, bottom left, and bottom right). c, Additional tracking examples: stingray track ending when the animal dived beneath the robot, barracuda track maintaining target lock despite specular reflections, and sustained sea turtle pursuit. d, Tracking pipeline: stereo cameras are stitched into a wide-field view, segmented to extract target centroid, which drives autonomous swimming control.
Animal Type
Track duration
Failed
End Track Reason
(secs)
Turtle
59
No
stopped swimming a
Turtle
57
No
stopped swimming a
Turtle
43
No
visibility
Turtle
38
No
coral collision b
Stingray
32
No
swam over robot
Table 1: Analyses of several aquarium tracks
Figure 3: Autonomous underwater obstacle avoidance. Top, Depth over time showing the robot maintaining a target depth of 1.1 m (RMSE = 0.195 m, MAE = 0.134 m). Bottom, Obstacle distance estimates over time with the 2.5 m avoidance threshold (dashed); yellow and blue shaded regions indicate right and left turn commands, respectively. Annotated time points highlight three representative avoidance events—coral reef (26 s), glass wall (68 s), and a diver (320 s)—each shown with the robot’s stereo camera view (Robot POV), the corresponding stereo depth map, and an external observer perspective.
Figure S1: System Diagram of Crush’s Autonomous Control and Tasking: the robot is equipped with cameras and several sensors, allowing it to operate autonomously with manual override and high-level tasking capability.
Figure S2: Robotic Sea Turtle Characterization: a) Image sequence showing the robot’s forward swimming trajectory in a pool environment. b) Underwater view of the robot during a diving maneuver, demonstrating flipper-based propulsion and pitch control. c) Motor kinematics during steady swimming: i) joint positions showing synchronized oscillatory flipper motion across all 10 motors (front and rear flippers), and ii) corresponding motor velocities demonstrating consistent stroke patterns. d) Performance characterization across N=5 trials: i) Forward swimming speed showing mean position ± 1 standard deviation over 30 seconds, ii) diving behavior showing depth excursion of approximately 1.5 body lengths with controlled ascent and descent phases, and iii) power consumption profile during steady swimming with mean power of 24.0 W yielding a cost of transport of 1.18 (dimensionless). The robot demonstrates repeatable locomotion with low trial-to-trial variability in both forward swimming and diving maneuvers.
Figure S3: Three untethered obstacle avoidance trials at the New England Aquarium (132 s, 94 s, and 341 s; left to right). For each trial, depth over time (top) and estimated obstacle distance (bottom) are shown, with yellow and blue shaded regions indicating right and left turn commands, respectively, and red shaded regions marking diver interference events. The robot maintained a 1.1 m target depth across all trials (RMSE = 0.213, 0.266, and 0.195 m; MAE = 0.135, 0.158, and 0.134 m) while executing 10, 7, and 28 avoidance maneuvers per trial, respectively. Across all three trials, the robot autonomously avoided 41 of 45 detected obstacles (91.1% success rate, 113.5 s per intervention), with the three wall-collision failures attributable to the glass tank environment rather than reef-relevant obstacles.
Figure S4: Untethered toy turtle tracking over a 32-second window of a 3 min 34 s trial. Top row: external observer view of the robot pursuing the toy turtle. Bottom row: onboard stereo camera view with the FOMO-detected centroid (red dot) used by the motion planner to drive the robot’s motor commands in real time.
Figure S5: Tetherless Tracking Trials. Top-down view of the autonomous tracking robot pursuing a toy turtle target across multiple trials in the MIT Sea Grant testing tank. Each semi-transparent silhouette represents the robot at successive positions during left, right, and center target trials. The glass wall at the center of the tank presented detection challenges, requiring the target to be repositioned closer to the robot for reliable model detection.
Underwater robotics develops autonomous and remotely operated machines for navigation, inspection, manipulation, and exploration in aquatic environments.
Increasing interest in deep-sea operations and resources motivates the development of ecologically sensitive but environmentally durable robots. Dielectric elastomer actuator artificial muscles are good candidates for powering such systems due to their pressure and temperature tolerance and soft makeup, but they are difficult to integrate with robotic systems. This work presents an autonomous robotic platform: the CORE, capable of driving six artificial muscles while sensing visual and spatial information. To validate the platform, we developed the Cuttlebot - a cuttlefish-inspired robot that swims in three dimensions using undulatory fin locomotion. The Cuttlebot has four primary artificial muscles in its fins in addition to a tentacle-inspired soft gripper. The robot was evaluated in a series of tethered and untethered swimming tests, demonstrating a top speed of 2.5 centimeters per second translation and 10 degrees per second rotation. Furthermore, the CORE system was capable of driving specialized control signals into the artificial muscles to controllably output force and torque in six axes. This work provides a platform for developing complex, bio-inspired swimming robots for ocean exploration and monitoring, laying the foundation with our leading example: the Cuttlebot.
Alexander Nicholas White, Ang Leo Li, Alexander Yin +5
School of Mechanical, Aerospace, and Manufacturing Engineering, University of Connecticut; Storrs, CT, U.S.A. · Department of Mechanical and Industrial Engineering, University of Toronto; Toronto, Ontario, Canada. · University of Connecticut, Institute of Materials Science; Storrs, CT, U.S.A. +1
Autonomous underwater vehicles (AUVs) are a promising alternative to divers for scalable collection of natural ocean ecology data. However, these robots may disturb local fauna and cause drastic behavioral differences compared to other monitoring techniques, decreasing their value as scientific tools. A promising prospect is to make AUVs that are more biomimetic, with the hope that taking on the form and behavior of a non-predatory animal may reduce adverse responses. We present the first dataset comparing fish disturbance in response to a conventional thruster-driven AUV, a sea turtle inspired flipper-driven AUV, and a diver. Experiments were conducted at a biodiversity hotspot in a Caribbean reef, and images of the scene were analyzed using computer vision to localize and study fish behavior change. Both AUVs cause measurable changes in fish behavior. Although no between-robot differences remain significant after correction for multiple comparisons, point estimates generally favor the biomimetic AUV, motivating larger studies capable of resolving modest effects and further design changes to optimize for disturbance. We also observe larger responses during diver transects than during AUV transects, although this exploratory comparison is based on a small diver sample with several confounds. While conclusions must be taken as preliminary due to operational and experimental limitations, this study provides the community with a first known dataset and benchmarks to quantify the behavioral impacts of biomimetic robots for ecological monitoring in the wild.
Oyster reefs are critical ecosystem species that sustain biodiversity, filter water, and protect coastlines, yet they continue to decline globally. Restoring these ecosystems requires regular underwater monitoring to assess reef health, a task that remains costly, hazardous, and limited when performed by human divers. Autonomous underwater vehicles (AUVs) offer a promising alternative, but existing AUVs rely on geometry-based navigation that cannot interpret scene semantics. Recent vision-language models (VLMs) enable semantic reasoning for intelligent exploration, but existing VLM-driven systems adopt an end-to-end paradigm, introducing three key limitations. First, these systems require the VLM to generate every navigation decision, forcing frequent waits for inference. Second, VLMs cannot model robot dynamics, causing collisions in cluttered environments. Third, limited self-correction allows small deviations to accumulate into large path errors. To address these limitations, we propose CORAL, a framework that decouples high-level semantic reasoning from low-level reactive control. The VLM provides high-level exploration guidance by selecting waypoints, while a dynamics-based planner handles low-level collision-free execution. A geometric verification module validates waypoints and triggers replanning when needed. Compared with the previous state-of-the-art, CORAL improves coverage by 14.28% percentage points, or 17.85% relatively, reduces collisions by 100%, and requires 57% fewer VLM calls.
Zhenqi Wu, Yuanjie Lu, Xuesu Xiao +1
University of South Florida · George Mason University