For the problem that the positioning algorithm has a large error in wireless sensor networks under different communication models, based on the average network connectivity of the off-line computation of Amorphous algorithm, four RSS (Received Signal Strength) threshold models are built up to suppress the positioning error of Amorphous algorithm under different communication models. The threshold obtained from different threshold models can be utilized to modify the gradient of the algorithm to different extents, so that the positioning error is suppressed. Simulation results show that, when Amorphous algorithm passes through the Regular model, the optimal threshold model is selected between Regular threshold model and Log-normal threshold model. When the algorithm passes through the Log-normal model, the optimal threshold model is selected among Regular threshold model, Log-normal threshold model, DOI threshold model and RIM threshold model. When the algorithm passes through the DOI model and RIM model, the optimal threshold model is selected among Log-normal threshold model, DOI threshold model and RIM threshold model. Finally, the optimal threshold model of Amorphous algorithm under different communication models, communication radiuses and irregularity degrees can be obtained.