Iteratively Re-weighted Least Squares (IRLS) were used to simulate the $L_p$-norm approximation of the ballistic trajectory in absolute gravimeters. Two iterations of the IRLS delivered sufficient accuracy of the approximation without a significant bias. The simulations were performed on different samplings and perturbations of the trajectory. For the platykurtic distributions of the perturbations, the $L_p$-approximation with $3<p<4$ was found to yield several times more precise gravity estimates compared to the standard least-squares. The simulation results were confirmed by processing real gravity observations performed at the excessive noise conditions.