NavySim: Multi-Vessel Simulation Engine

Jun 24, 2026 · 2 min read
NavySim 2.0 system architecture and live multi-vessel tactical simulation with CPA threatmaps

NavySim: Multi-Vessel Simulation Engine (2022–2026) is a serious-game simulation and analysis platform built in Unity for naval autonomy research, developed during my doctoral studies at the University of Nevada, Reno.

System Architecture & Key Capabilities

  • Modular System Architecture: Features decoupled subsystem managers (Scenario Manager, Agent Manager, Physics Engine, Threat Evaluation, and Sensor/Weapon Managers) facilitating rapid scenario prototyping and benchmarking.
  • Physics-Consistent Multi-Agent Scenarios: Supports heterogeneous surface vessels with configurable hydrodynamics, COLREGS-compliant collision avoidance (VOCCA), sensing envelopes, and defensive suites.
  • Dynamic Threat & Vulnerability Surfaces: Real-time shader pipeline calculates Closest Point of Approach (CPA/TCPA), sensor blindness zones, and weapon coverage to render unified threat surfaces.
  • Integrated Machine Learning Loop: Connects external PyTorch and statistical intent models (HMMs, LSTMs, Transformers, MTITP GANs) via high-throughput TCP streaming for real-time tactical decision support.
  • Industry & Academic Deployment: Validated in collaboration with naval researchers and industry partners (Huntington Ingalls Industries) for autonomous vessel interaction analysis.

NavySim 2.0 Modular System Architecture
Figure: High-level modular architecture of NavySim 2.0 (IEEE Transactions on Games 2026), detailing decoupled Scenario, Physics, Agent, and Threat Evaluation subsystems with high-throughput ML bridge.

Multi-Vessel Tactical Simulation in Unity
Figure: Multi-vessel tactical encounter in NavySim demonstrating physics-based ship maneuver dynamics, sensor coverage, and live threat calculation.

Simulation Engine Maritime AI Unity / C# Multi-Agent Systems System Architecture
Md Abu Sayed
Authors
Md Abu Sayed
Postdoctoral Researcher
Cardiovascular AI & CDSS · Joint Intent Recognition & Trajectory Prediction · Deep Learning