Technical Challenge
The integration of heterogeneous unmanned underwater, surface and aerial systems into cohesive, adaptive maritime fleets represents one of the most demanding challenges in modern undersea defence. Developing and validating the algorithms needed for distributed adaptive autonomy, multi-domain sensor fusion, and effective human-machine teaming requires a simulation environment capable of faithfully reproducing the physical, sensory, and operational complexity of real maritime scenarios. Currently, no open, modular, and scalable simulation testbed exists that can simultaneously support all three operational domains, multi-agent coordination at fleet scale, and human-in-the-loop experimentation, creating a critical bottleneck for both academic research and industrial R&D.
Methodology
Naval Group, with the IRL CROSSING International Research Laboratory (a joint Australia–France initiative), has developed LOTUSim (Learning from Operational Teaming with Unmanned Systems), an open-source, modular, distributed simulation platform designed to serve as a comprehensive testbed for maritime multi-agent research. LOTUSim is built on a layered architecture combining Gazebo as its simulation core, ROS 2 with FastDDS as a distributed communication middleware, and xdyn (Sirehna) as its hydrodynamic physics engine. Its design philosophy rests on two pillars: radical modularity — any subsystem can be independently enabled, replaced, or extended — and distributed execution, whereby computational loads are offloaded to external engines running on dedicated hardware.
LOTUSim is deployed as the integration and validation platform for the maritime programme of IRL CROSSING. It provides the shared simulation environment through which distributed adaptive autonomy, multi-modal event recognition and situation narration, and adaptive human-machine interfaces are jointly validated before real world trials. PhD candidates across IRL CROSSING use LOTUSim as their primary research infrastructure, with results including sim-to-real transfer of deep reinforcement learning controllers for AUV fault recovery, multi-agent benchmarking, and narrative generation from autonomous multi-robot missions.
Key Findings
LOTUSim has demonstrated its viability as a multi-domain simulation backbone for maritime autonomy research. Early results from IRL CROSSING doctoral programmes show successful sim-to-real transfer of adaptive AUV control algorithms, validating the fidelity of the platform for reinforcement learning. The open-source, EPL 2.0 licensing model has enabled contributions from a global community, accelerating the development of reference implementations for dynamics, sensors, and multi-agent management. The platform's modular architecture has allowed rapid prototyping of human-machine teaming scenarios, a critical enabler for man-machine teaming and operator cognitive load. The 2026 roadmap, spanning sensor subsystems (camera, AIS, sonar, radar), power management, and AI/RL simulation campaigns, directly anticipates the operational requirements identified.
LOTUSim represents a paradigm shift in how the maritime autonomy community can conduct simulation-driven research: from isolated, purpose-built simulators to a shared, open, industrially grounded platform maintained by Naval Group and the IRL CROSSING international laboratory. By providing a scalable testbed, supporting deterministic and stochastic execution, and offering clean interfaces for autonomy stack, and human-machine interfaces, LOTUSim directly addresses collaborative autonomy, multi-platform coordination, AI-driven decision support and man-machine teaming. The platform's role as a validation backbone positions it as a strategic enabler for the next generation of undersea and surface defence capabilities, with a clear pathway from simulation validation to open-sea experimental trials.