Localizing Unsynchronized Sensors With Unknown Sources

We propose a method for sensor array self-localization using set of sources at unknown locations. The produce signals whose times arrival are registered the sensors. look general case where neither emission nor reference time frames receivers known. Unlike previous work, our directly recovers geometry, instead first estimating timing information. key component is new loss function which insensitive to timings. cast problem as minimization non-convex functional Euclidean distance matrix microphones and subject certain constraints. After convexification, we obtain semidefinite relaxation gives an approximate solution; subsequent refinement on proposed via Levenberg-Marquardt scheme final Our achieves state-of-the-art performance in terms reconstruction accuracy, speed, ability work with small number receivers. It can also handle missing measurements exploit prior geometric temporal knowledge, example if either receiver offsets or known, contains compact subarrays known geometry.