Core Concepts
Jointly solving multi-source localization and data association using optimal transport.
Abstract
In this work, the authors address the challenge of localizing multiple signal sources based on time-difference of arrival (TDOA) measurements in a blind setting. They propose a method that combines localization and data association by utilizing an optimal transport formulation. The approach involves finding optimal groupings of TDOA measurements and associating them with candidate source locations. By constructing an efficient set of candidate locations using minimal multilateration solvers, the proposed method demonstrates robustness to measurement noise and TDOA detection errors in three-dimensional space. The joint solution allows for statistically efficient estimates of source locations, overcoming challenges posed by unknown source signals and data labels.
Stats
In numerical simulations, we demonstrate that the proposed method is robust both to measurement noise and TDOA detection errors.
For each pair, create all combinations of TDOAs, with one from each pair.
The relative source-receiver distances are then encoded in TDOAs corresponding to a source-receiver-receiver triplet.
The number of minimal sets of receiver pairs was set to K = 3.
The standard deviation σ of the error was varied between [0.01, 0.19].
Quotes
"The proposed method displays considerable robustness to TDOA noise."
"The obtained association allows for a refinement stage achieving statistical efficiency."
"The joint solution allows for statistically efficient estimates of source locations."