2025_programme: A Submodular Approach to Acoustic Sensor Placement using the MARLIN Digital Twin

• Day: June 19, Thursday
    Location / Time: A. TERPSIHORI at 12:00-12:20
• Last minutes changes: -
• Session: 12. Observing the Oceans Acoustically
  Organiser(s): Bruce Howe, Kay Gemba
  Chairperson(s): Bruce Howe, Kay Gemba
• Lecture: A Submodular Approach to Acoustic Sensor Placement using the MARLIN Digital Twin
  Paper ID: 2293
  Author(s): Edward Clark, Eric Schoof, Alan Hunter, Chris Manzie
  Presenter: Edward Clark
  Abstract: We present a greedy algorithm for acoustic sensor placement within the MARLIN digital twin. \nIn this study, we aim to optimize the placement of N sensors in a test region off the coast of Victoria, Australia to maximize target detection probability using a greedy algorithm.\n\nWe postulate that the sensor placement problem can be formulated as a submodular optimisation problem. This is where adding a sensor to a set of sensors will have a diminishing but always increasing return on the objective function. This submodular approach ensures a greedy algorithm will perform within 63\% of the optimal solution.\n\nMARLIN is a digital twin that simulates real-world ocean acoustic environments, utilizing open-source data from the Copernicus Marine Toolbox and GEBCO to create high-fidelity models. Our dataset comprises over 2,000 simulated sensors along a 70 km transect, uniformly distributed from 0–2 km depth.\n\nFor increasing numbers of sensors placed we calculate the region-wide average detection probability, minimum detection probability, and maximum detection probability.\nMonte Carlo simulations are used to evaluate the random placement algorithm and compare its performance with the greedy algorithm. Results demonstrate that the greedy algorithm consistently outperforms a random approach, providing a stronger lower bound for reinforcement learning agents often benchmarked against random baselines.\nWe also examine the effect of varying noise levels on algorithm performance, demonstrating their adaptability to different acoustic conditions. This study highlights the potential of the greedy algorithm as an effective method for sensor placement in complex marine environments, offering insights for future research and practical applications.\n
    Download the full paper
  This paper is a candidate for the "Prof. Leif Bjørnø Best Student Paper Award (for students under 35)"
• Corresponding author: Mr Edward Clark
  Affiliation: University of Bath
  Country: United Kingdom