Seismic Imaging of Lunar Lava Tubes using Optimal Adaptive Sensor Deployment: Results from a Terrestrial Analog Campaign

Kurzfassung

This study investigates adaptive seismic sensor deployment for detecting and characterizing lunar lava tubes under realistic mission constraints. Because the imaging quality of subsurface voids depends strongly on sensor geometry, the authors develop a Bayesian Optimal Experimental Design framework in which a pre-trained neural network proposes sensor locations sequentially. Later placements adapt to measurements acquired earlier in the survey, allowing the network configuration to evolve with the available information. Operational constraints are incorporated directly into the design process: communication connectivity is enforced through prediction of a connected tree structure, while localization performance is represented using the Cramer-Rao lower bound derived from inter-agent radio-frequency ranging geometry. Synthetic lunar simulations show that the adaptive strategy outperforms non-adaptive and heuristic deployment schemes while respecting these constraints. The method is further evaluated through a terrestrial analogue campaign targeting the La Corona lava tube system on Lanzarote, Spain. There, seismic ambient-noise measurements are collected using sensor positions proposed by the neural network to assess the feasibility of autonomous, constraint-aware mapping of lunar subsurface voids.