LEMURS
Project reference: CNS2024-154526
Start date: 01/04/2025
End date: 31/03/2027
Coordinated by: Universitat de Girona
The LEMURS project is developing a new generation of Intervention Autonomous Underwater Vehicles (I-AUVs) capable of performing complex scientific tasks independently. Our mission is to reduce reliance on costly, ship-based Remotely Operated Vehicle (ROV) operations by enabling efficient and intelligent autonomous systems that contribute to the protection of marine biodiversity.
Analyze current ROV pilot and scientist manipulation practices to define automation challenges and identify effective learning-based solutions.
Build a high-fidelity Girona500 I-AUV simulation environment to enable accelerated training, benchmarking, and sim-to-real transfer.
Design and benchmark a hybrid DRL + Imitation Learning architecture against expert systems and human demonstrations for robust underwater manipulation.
Our methodology follows a structured four-stage pipeline to enable a reliable transition from simulation to real ocean conditions. First, idealized tasks would be solved in a perfect virtual environment using privileged, noise-free data. The robot would then be restricted to onboard sensing only, such as vision and tactile feedback. Next, noise and domain randomization would be introduced to improve robustness against real-world uncertainty. Finally, real trajectory data would be incorporated and the learned policies transferred to the physical Girona 500 I-AUV for field deployment.
Reinforcement Learning (RL)
Learning by Demonstration (LbD)
Human-in-the-loop training
Intervention AUV (Girona 500)
Advanced perception systems
Force and tactile sensing
Real-world testing in large water tanks
Sim-to-real transfer
One envisioned application of LEMURS is the autonomous deployment of underwater sensor networks. An intervention AUV can detect, grasp, transport, and precisely place hydrophones on the seabed, forming optimal monitoring grids without the need for human intervention or large vessels.
Learning a robotic manipulation task often involves several actions that must be completed in the right order. This challenge becomes even more important underwater, where
One of the key challenges in the development of the LEMURS project is the availability of an accurate and efficient simulation environment. Such a simulator
Within the framework of the LEMURS project, Task T2.2 – ML-based Algorithm Evaluation for Manipulation has been successfully carried out through two Master’s theses conducted
Honored to have been a speaker at RobSummit (2nd Edition) at the University of Porto, invited by our colleagues at INESC TEC. Following insightful sessions
UrbanAcres, LLC
“My move was way less stressful than expected! Movely packed everything, stored it for a week, and moved it all to my new home just as promised!”
H&H Legal Partners
“Movely provided a very professional service when we moved offices last month. They were easy to work with and moved everything on time.”
SavvyMarket
Best experience ever! I desperately needed a mover to help with a last minute relocation, and Movely was so helpful in making it a seamless process!”
© 2026 Movely. All Rights Reserved.