Human-machine teaming dives underwater

2026-04-14 · Source: MIT News - Artificial intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems · Depth: Advanced, long

Summary

MIT Lincoln Laboratory researchers are developing hardware and algorithms to enhance human-machine teaming for underwater missions, specifically between divers and Autonomous Underwater Vehicles (AUVs). This project, funded by an internal R&D portfolio on autonomous systems, aims to combine human dexterity and object recognition with robot processing power, high-speed mobility, and endurance. Key developments include navigation algorithms that account for ocean currents, requiring more sophisticated diver sensing, and an AI classifier that processes both optical and sonar data mid-mission. This classifier solicits human input for uncertain object classifications, necessitating efficient underwater acoustic communication despite low bandwidth and high latency. The team has integrated COTS sensors into a U.S. Navy-compatible AUV payload and conducted field tests in coastal New England and the Great Lakes, using diver surrogates and human divers with a prototype "tube-let" tablet.

Key takeaway

For Computer Vision Engineers developing underwater systems, you should prioritize robust navigation and perception solutions that account for dynamic ocean conditions. Focus on integrating multi-modal sensing (optical and sonar) with AI classifiers that can solicit real-time human feedback, optimizing data compression for low-bandwidth acoustic communication. This approach will enhance operational effectiveness and safety for divers in complex undersea environments.

Key insights

Combining human and robot strengths is crucial for effective underwater human-robot teaming in challenging environments.

Principles

• Humans excel at dexterity and object recognition underwater.
• Robots offer superior processing, speed, and endurance.
• Real-world ocean forces complicate underwater navigation algorithms.

Method

The team developed navigation algorithms and an AI classifier for optical/sonar data, integrating COTS sensors into an AUV. They also explored knowledge transfer from optical to sonar classifiers to reduce data labeling burden.

In practice

• Use AI classifiers with human feedback for uncertain object identification.
• Employ acoustic modems for diver-AUV communication.
• Integrate COTS sensors for rapid technology transition.

Topics

• Human-Robot Teaming
• Autonomous Underwater Vehicles
• Underwater Navigation
• Underwater Perception
• Sonar Data Analysis

Best for: Computer Vision Engineer, AI Scientist, Robotics Engineer, Research Scientist

Related on AIssential

• Featured video: Coding for underwater robotics — Underwater navigation algorithms can enable collaborative human-robot operations despite GPS limitations.
• Drones paired with AI could help search-and-rescue teams find missing persons faster — AI-enhanced drones with multi-spectral imaging can significantly improve wilderness search-and-rescue efficiency and victim assessment.
• New imaging system sees through murky waters — Opti-acoustic fusion enables real-time, high-resolution 3D underwater mapping in turbid conditions by combining sonar's scale with optical detail.
• Table tennis robot defeats some of world’s best players – why this has major implications for robotics — Ace robot demonstrates AI's ability to achieve superhuman performance in complex, real-world physical environments.
• Task-specific Subnetwork Discovery in Reinforcement Learning for Autonomous Underwater Navigation — Multi-task RL networks achieve task differentiation with minimal, context-driven weight specialization.
• Improving Human-Robot Teamwork in Urban Search and Rescue Through Episodic Memory of Prior Collaboration — Robots can significantly improve human-robot teamwork by leveraging human-externalized collaboration patterns as episodic memory.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by MIT News - Artificial intelligence.