LocoMote: AI-driven Sensor Tags for Fine-Grained Undersea Localization and Sensing

Area of Science:

• Marine technology
• Underwater sensor networks
• Artificial Intelligence in oceanography

Background:

• Maritime localization and sensing face challenges like intermittent connectivity, power limits, and harsh environments.
• Existing technologies struggle with long-term, fine-grained underwater data collection.

Purpose of the Study:

• To design and implement LocoMote, a rugged, ultra-low-footprint undersea sensor tag.
• To enable on-device AI-driven localization, online communication, and energy harvesting for underwater applications.

Main Methods:

• Utilized on-chip neural networks (< 30 kB) for AI-driven localization using inertial sensor data.
• Implemented piezo-acoustic ultrasonics for underwater data streaming (2-5 kbps) with a 50m+ range.
• Integrated an aluminum-air salt water energy harvesting system for battery recharging (up to 5 mW).

Main Results:

• Achieved underwater object tracking within 3 meters during ~6 minutes of GPS outage.
• Enabled concurrent data streaming from up to 55 nodes using randomized time-division multiple access.
• Developed an ultra-lightweight (< 50g), compact tag with low power consumption (~330 mW peak).

Conclusions:

• LocoMote offers a viable solution for long-term, fine-grained maritime localization and sensing.
• The tag's integrated AI, communication, and energy harvesting address key limitations of current underwater technologies.
• Demonstrated real-world performance superior to existing oceanic sensing technologies.