Optimizing Marine Sensors to Capture Detailed Ocean Data Efficiently
ECE Associate Professor Milad Siami, in collaboration with the Woods Hole Oceanographic Institution (WHOI), is leading a $612,062 NSF grant for “Energy-Aware Deployment and Scheduling for Sparse Ocean Sensor Networks.” The project builds on work done in Siami’s Lab in multi-agent systems, network science, systems and control, and robotics. Together with WHOI oceanographers, the team will develop mathematical tools to decide where drifting sensors should go and when they should transmit, so that fewer devices and less power can reveal more of the sea’s hidden currents. The partnership pairs Northeastern’s expertise in network optimization and control theory with WHOI’s world-class knowledge of ocean physics and observation.
Abstract Source: NSF
This project aims to make ocean monitoring more efficient by rethinking how floating sensors are deployed and operated. Scientists rely on drifting ocean sensors (“drifters”), which periodically report their positions, to track currents that redistribute heat, pollution, and marine life. However, drifters are costly and run on limited batteries, and their deployment locations and reporting schedules are set by rule of thumb—wasting energy in some regions while missing important activity in others. The project will bring transformative change by treating a fleet of drifters as a connected network whose overall structure can be optimized. Mathematical tools will be developed for deciding where to release sensors and when each one should transmit, so that the fleet reveals ocean features using fewer devices and less power. The intellectual merit of the project includes new theory linking network science, control engineering, and ocean physics, with provable guarantees that scientific accuracy is preserved under tight energy budgets. Broader impacts include improved oil-spill and search-and-rescue response, longer and more efficient climate observation records, training of students across engineering and oceanography, new publicly available course materials, public outreach, and opensource software.
The project models an oceanic sensor fleet as a dynamic time-evolving “similarity graph,” in which each sensor, or drifter, is a node, and connections (edges) reflect how similarly two drifters move. The mathematics of such graphs, captured by the graph Laplacian, encodes the ocean’s key transport structures (eddies or jets) and their resulting transport pathways and barriers. Three central challenges will be addressed. First, developing optimal deployment strategies to recover transport structures within a fixed budget, while accounting for sensor losses and transmission gaps due to unpredictable events. Second, developing energy-aware scheduling “sleep” of redundant sensors to save power. These tasks will use a structure-preserving model reduction technique called Kron reduction, optimally maintaining the fleet’s information content when nodes are removed. The core technical challenge is guaranteeing that removed or resting sensors do not corrupt detected ocean features. The project will resolve this challenge by estimating error bounds that to maintain clustering accuracy, measured by the Adjusted Rand Index (a standard score for comparing groupings). Third, the methods will be validated on idealized flows, realistic ocean-model simulations, and large historical datasets, and released as open-source software for operational use. This work will establish a rigorous, energy-aware standard for sparse ocean sensing and a bridge between control theory, network science, and physical oceanography.