Exploring H-SURF, SUAD Project Awarded Sheikh Hamdan bin Zayed Environmental Research Award

Collective Intelligence under the sea

The exploration and preservation of marine environments demand autonomous systems capable of long-term mission, adaptability to changes of the environment, and resilience to possible damages and misfunctioning. Fleets of autonomous robots coordinating for collaborative missions and task allocation represent a transformative solution for the future of environmental preservation. However, applications of underwater multi-robot fleets require robust coordination strategies that often stress the limited capabilities of underwater vision, localization and communication.

Dr. Giulia De Masi, Associate professor at Sorbonne University Abu Dhabi, explains the Research Project ⁠"H-SURF: Heterogeneous swarm of underwater robotic fish: from biological fish swarms to artificial fish swarms and back" which has been awarded by the Sheikh Hamdan bin Zayed Environmental Research Award, Environmental Research Project category.

Figure: Sheikh Hamdan bin Zayed Environmental Research Award Medal, a important milestone for the SDG 14 and 13 of United Nations

The H-SURF is a bio-inspired collective system designed to overcome the limitations of traditional Autonomous Underwater Vehicles (AUVs) in complex, dynamic ecosystems. Drawing inspiration from natural schooling of biological fish, the robot fleet minimizes ecological disruption, while its modular design allows for rapid reconfiguration for missions ranging from environmental monitoring to observation of fish species.

This is a fully realized system designed to overcome the underwater challenges through bio-inspired, decentralized control and Artificial Intelligence on the edge.

Figure: two elements of the school of H-SURF robotic fish [1]

The system architecture is uniquely heterogeneous, consisting of a school of artificial robotic fish supported by two auxiliary platforms: a floater to guarantee connectivity with land-based research stations, and a sinker to provide underwater telepresence and global monitoring.

The methodologies have been validated across three stages: numerical simulations; lab-scale experiments for sensor-actuator validation; and open-water deployments in UAE coastal environments, proving the system's translational applicability under real ecological constraints.

Individual and Collective Artificial Intelligence

To manage coordination, a multilayered control architecture leverages visual recognition and swarm behaviors, implementing individual and collective Artificial Intelligence.

Figure: A representation of the different layers, and the types of localization and communication strategies implemented at each level. Starting from single layer, with only basic sensing, local group dominated by visual, super-group depending acoustic, user layer interfacing from the surface, and floater sinker acting as bridge among the different layers [2]

The onboard AI enhanced vision system of the single fish has been already applied to detect dolphins in UAE offshore in collaboration the UAE Dolphin Project.

Figure: Underwater species recognition based on Computer Vision Deep learning models [3]

Homegrown in UAE (2020-24), funded by TII, and developed at Khalifa University, the three PIs, Dr. Giulia De Masi (Sorbonne University Abu Dhabi), Federico Renda (Khalifa University) and Cesare Stefanini (MBZUAI) are currently based in UAE, keeping a fruitful collaboration across UAE institutions.

Figure: the 3 PIs of the project: Dr. Giulia De Masi (Sorbonne University Abu Dhabi), Federico Renda (Khalifa University) and Cesare Stefanini (MBZUAI)

Next steps

Dr. Giulia De Masi and her team continue to develop AI algorithms for species detection and biosystem monitoring of the underwater environment, and coordination of autonomous vehicles underwater, in particular developing models of Multi-Agents Reinforcement Learning, an important field of Artificial Intelligence.

This research investigates the emergence of collective intelligence within swarms of autonomous underwater vehicles, drawing direct inspiration from the cooperative behaviors seen in marine life. At the heart of this system is Reinforcement Learning, a computational framework where each robot refines its navigational and exploratory skills through a process of trial and error, mirroring the way humans and animals learn from their environment. Rather than relying on rigid, pre-set instructions, these agents develop their own internal logic to navigate the complexities of the sea.

As these robots interact, they move beyond individual strategy to discover sophisticated strategies for teamwork. This adaptability is crucial when environmental conditions shift unexpectedly. Even though each robot operates based only on its limited local observations, the collective ability to learn and distribute information allows the entire group to overcome uncertainty typical of the underwater environment and achieve a common mission of monitoring and inspection.

References

1) Enhancing Collaboration in Uncertain Environment: Multi-Agent Reinforcement Learning for Underwater Monitoring, Alberto Luvisutto, Antonio Celani, Federico Renda, Cesare Stefanini, and Giulia De Masi, Expert Systems With Applications, Volume 277, 5 June 2025, 127256 (DOI: 10.1016/j.eswa.2025.127256)

2) Heterogeneous Underwater Swarm of Robotic Fish: Behaviour and Applications, S. Iacoponi, M. Hanbaly, A. Infanti, B. Andonovski, N. Mankovskii, I. Zhilin, F. Renda, C. Stefanini, G. De Masi, Embodied Intelligence Conference 2022, IOP Conference Series: Materials Science and Engineering 1292 (1), 012008, IoP Press , 2023

3) Underwater Inspection Platform for Vision-Based Biodiversity Identification, Gianluca Manduca, Gaspare Santaera, Ada Natoli, Giulia De Masi, Cesare Stefanini, Donato Romano, 2025 IEEE International Workshop on Metrology for the Sea; Learning to Measure Sea Health Parameters (MetroSea), Genoa, Italy, 2025, pp. 45-49.