Industrial Consultancy & Sponsored Research (IC&SR) , IIT Madras
A System And Method For Autonomous Marine Navigation
Technology Category/Market
Category- Artificial Intelligence & Machine Learning
Industry Classification:
Shipping, Defense, Surveillance and Security, Ports, Artificial Intelligence
Applications:
Autonomous navigation of surface marine vehicles such as ships, ASVs, and boats; Retrofit into existing manned vessels to enhance autonomy; Training simulators for maritime navigation systems.
Market Drivers
The Global The global autonomous navigation market was valued at USD 3.5 billion in 2024 and is projected to grow to USD 9.4 billion by 2032, with a CAGR of 13.1%.
Problem Statement
- Autonomous marine navigation enhances safety, reduces human error, and increases operational efficiency in applications like shipping, defense, research, and environmental monitoring.
- Conventional navigation systems use static rule-based algorithms, GPS, and basic sensors that lack adaptability to complex, dynamic, or hazardous marine environments.
- Further, these systems do not integrate learning-based adaptability, have poor response to unpredictable obstacles, and lack continuous improvement or modular integration capabilities.
- There is a need for a deep reinforcement learning model that integrates with real-time sensors and adaptive modules for safer, smarter, and energy-efficient autonomous marine navigation.
Technology
- The system leverages a 4th-order Runge-Kutta MMG ship dynamics model with KVLCC2 parameters, ensuring high-fidelity simulation of surge, sway, and yaw motions.
- A Line-of-Sight (LoS) module continuously computes cross-track (CTE) and heading errors (HE), placing waypoints at twice the vessel length as look-ahead for anticipatory steering
- The TD3 training module employs an 8-400-300-1 feed-forward neural network, Adam optimizer (LR 0.0005), batch size 100, and ε-greedy exploration (ε: 0.999→0.1) for sample-efficient, stable policy learning
- Reward module integrates path following rewards, collision avoidance rewards and COLREGs rewards into a compound signal, driving balanced optimization of accuracy and safety
- Seamless system integration via standard protocols—UART, SPI, I2C, Bluetooth, Wi-Fi, LTE, TCP/IP—enables retrofit onto existing ASV, ship, and boat platforms without specialized hardware modifications
Key Features/Value Proposition
- Uses Twin Delayed Deep Deterministic Policy Gradient (TD3) module with continuous learning and dynamic policy refinement, unlike fixed rule-based systems.
- Integrates path following and obstacle avoidance in one seamless unit, unlike many systems focusing on single-function modules.
- Incorporates COLREGs and TCPA logic, improving decision-making during dynamic encounters far better than traditional reactive methods..
- Reduces rudder effort using intelligent decision-making; essential for long missions and ASVs with limited energy budgets.
- Easily integrable with existing marine vehicles due to modular hardware/software design and standard communication protocols.
Questions about this Technology?
Contact for Licensing
Research Lab
Prof. Suresh Rajendran
Prof. Abhilash Sharma Somayajula
Department of Ocean Engineering
Intellectual Property
- IITM IDF Ref 2875
- IN 202441025682 Patent Application
Technology Readiness Level
TRL 2
Technology Concept Formulated