A Modified Near-Field Gaussian Plume Inversion Method Using Multi-sUAS for Emission Quantification

This work introduces a modified near-field Gaussian plume inversion method (mod-NGI), for emission quantification from a point source using downwind vertical plane measurements. The original NGI method aims to solve the inverse problem by estimating model parameters using measured data and a partial parameter least squares fitting. Given noisy measurements, this can lead to ill-posed conditions, poor performance, and sensitivities to initial conditions. The mod-NGI method combines preconditioned maximum likelihood parameter estimation with full parameter least squares optimization. The single sensor performance of mod-NGI method is explored with static and dynamic ground truth functions for fixed path (lawn mower and spiral) and adaptive path (extremum seeking control) planning strategies including a comparison with experimental NGI results. A high level multi-sUAS control strategy is developed for continuous parameter estimation using centroidal Voronoi tesselations and mod-NGI.