A review and case study on Underwater Legged Robot simulation

Underwater Legged Robots (ULRs) offer enhanced stability and precision for seabed mobility. However, translating these systems to real-world deployment requires high-fidelity virtual environments to develop and validate control and autonomy algorithms. To address this challenge, this research presents a comprehensive review of robotic simulators from the past decade. The analysis evaluates these platforms based on their hydrodynamics, seabed interaction, vehicle and sensor modeling, and Artificial Intelligence (AI) integration. Furthermore, it identifies critical gaps in the current state of the art that hinder the effective simulation of ULRs. Complementing this review, this paper benchmarks three simulation architectures through a unified hexapod ULR case study: the general-purpose Gazebo Sim, the marine-specific Stonefish, and the AI-driven NVIDIA Isaac Sim. To establish a consistent baseline for this comparison, a custom, GPU-accelerated hydrodynamics plugin was implemented within NVIDIA Isaac Sim to resolve its lack of baseline marine physics. The review reveals that current marine simulation platforms focus primarily on surface and water-column operations. Consequently, they often fail to integrate the multi-contact terrestrial mechanics and benthic hydrodynamics required for advanced ULR locomotion. The comparative analysis further demonstrates that, while framework selection ultimately depends on specific research objectives, successfully integrating multibody marine physics within parallelized architectures provides the computational foundation necessary to optimize AI-driven bio-inspired ULR locomotion.