Geophysical methods are considered as the most important and efficient methods for mineral exploration. Among them, the gravity method is widely used in the early stages of oil and gas as well as salt domes exploration. Due to inherent limitations such as inadequate number of data, noise and device limitations, the 2D 3D geologically consistent interpretation and modeling of gravity data is often necessary. Inverse modeling is a numerical tool for obtaining 2D 3D images of subsurface structures using measured data. In current study, a modified downhill simplex method is used to invert gravity data collected from Rudan region in the Hormozgan province. This method is a non-derivative geometric method that works with the objective function values at the vertices of the simplex polygonal. In order to model the negative anomaly of the probable hidden salt dome at the west of the Rudan region, the Talvani's 2D polygonal forward modeling approach was used in current study. A number of 6 profiles encompassing the low Bouguer anomaly at the west part of the study area were selected as the inputs for the inverse modeling program written in MATLAB environment. For this purpose, the response of individual and hybrid 2D polygonal models with different 2D polygonal cross sectional shapes was computed and compared with the residual anomaly after removing the third-order regional trend. In inverse modeling, the results of the modified downhill simplex method are compared with other non-gradient inversion methods such as genetic algorithm by applying upper and lower bounds on model parameters. It was found that although the RMS error for the genetic and simplex algorithms were 0.08 and 0.33 respectively, but the run time for the recovered final model in the genetic algorithm is much longer. However, the performance of the downhill simplex method is better than the genetic algorithm in terms of number of iterations in inverse modeling. It should also be noted that although the global optimization algorithms such as those employed in current study, are capable of ultimately finding the minimum of the objective function through searching in the model parameter space in a longer time, but through applying different constraints, a real and geologically consistent and valid results was obtained using downhill simplex approach.