What Is Physical AI?
Module 1 · Introduction to Physical AI in Maritime

What Is Physical AI?

Understand how AI moves beyond screens into the physical world — sensing, deciding, and acting in real environments like ships, ports, and oceans.

From Software AI to Physical AI

Most AI you interact with daily — chatbots, recommendation engines, image classifiers — lives entirely in software. It processes data and returns data. Nothing physically moves.

Physical AI is different. It closes the loop between the digital and physical worlds:

  1. Sense — cameras, radar, sonar, GPS, and other sensors feed real-world data into the system
  2. Perceive — AI models interpret that raw data (obstacle ahead, sea state rough, engine temperature rising)
  3. Decide — the system plans the best response given constraints (safety rules, mission goals, regulations)
  4. Act — motors turn, valves open, alarms sound, course corrections execute

This sense-perceive-decide-act cycle runs continuously — often hundreds of times per second — while the physical world changes around it.


Why Maritime Is a Prime Domain for Physical AI

The maritime industry is one of the most demanding environments for Physical AI for several reasons:

ChallengeWhy It’s Hard
Vast, unstructured environmentOpen ocean has no road markings, no traffic lights, no fixed lanes
Extreme sensor conditionsSaltwater, fog, spray, and darkness degrade cameras and lidar
Safety stakesA mistake in a cargo corridor can sink a vessel or cause an oil spill
Regulatory complexityCOLREGs, SOLAS, flag-state rules, and port authority requirements all apply
Crew limitationsFatigue, small crews on large vessels, and remote operations demand automation

Despite — or because of — these challenges, shipping carries 90% of world trade, and the industry is investing heavily in AI to improve safety, efficiency, and sustainability.


The Physical AI Stack in Maritime

Think of maritime Physical AI as five stacked layers, each depending on the one below it:

┌─────────────────────────────────────┐
│  5. Mission Layer                   │  voyage planning, fleet management
├─────────────────────────────────────┤
│  4. Decision Layer                  │  collision avoidance, route optimization
├─────────────────────────────────────┤
│  3. Perception Layer                │  object detection, sea-state estimation
├─────────────────────────────────────┤
│  2. Sensor Layer                    │  radar, AIS, cameras, sonar, IMU
├─────────────────────────────────────┤
│  1. Physical Layer                  │  hull, propulsion, rudder, actuators
└─────────────────────────────────────┘

Each layer feeds information upward and receives commands downward. A breakdown at any layer affects every layer above it. In Module 2 we explore the sensor layer in depth; in Module 3 we cover the decision layer.


Real-World Physical AI in Maritime Today

These systems are already operating commercially:

Autonomous Surface Vessels (ASVs) Rolls-Royce and Kongsberg have demonstrated remotely operated and semi-autonomous ferries in Norwegian fjords. The Yara Birkeland — a zero-emission container ship — sailed autonomously between Norwegian ports in 2022.

Predictive Maintenance Wärtsilä’s AI platform monitors engine vibration, exhaust temperatures, and fuel flow in real time. It predicts failures days before they happen, allowing planned maintenance instead of emergency repairs at sea.

Port Automation The Port of Rotterdam uses AI to coordinate berth scheduling, crane sequencing, and truck flow. Computer vision tracks every container in the yard.

Weather Routing AI Systems like StormGeo and Weathernews analyze satellite data and model forecasts to find fuel-optimal routes that avoid heavy weather — saving 5–15% on fuel per voyage.


💻 Code Lab: Key Terminology
TermDefinition
ASVAutonomous Surface Vessel — an unmanned boat operating on the water’s surface
AUVAutonomous Underwater Vehicle — an unmanned submarine-style vehicle
COLREGsInternational Regulations for Preventing Collisions at Sea — the “rules of the road” for ships
IMUInertial Measurement Unit — measures acceleration and rotation; used for navigation
AISAutomatic Identification System — vessels broadcast their position, speed, and identity via AIS
COLAVCollision Avoidance — the onboard system that detects and responds to collision risk

🧠 Knowledge Check

What distinguishes Physical AI from traditional software AI?

Reveal Answer & Explanation

Physical AI closes the sense-perceive-decide-act loop in the real world. Unlike pure software AI that takes data in and returns data out, Physical AI has actuators that change the physical environment — turning a rudder, sounding an alarm, or stopping an engine. The consequences of errors are physical and often irreversible, which is why safety and reliability requirements are far stricter than for software-only systems.


What’s Next

In Lesson 2, you’ll explore the maritime sensor ecosystem — radar, AIS, lidar, cameras, and sonar — and learn how data from these sensors is fused together to give an AI system a coherent picture of its surroundings.