A finite element code ABAQUS/Explicit was used to model three-dimensional soil compaction process. Drucker-Prager Cap (DPC) model was used as soil constitutive model. The behavior of the soil bionic press roller interface was investigated and compared with a conventional roller. Simulation results show that the artificial strain energy is approximately 0.1% of the internal energy, indicating that hourglassing exerted very little influence on the simulation results. For the conventional press roller and bionic press roller, the relative error of traction resistance between the simulation and soil bin test results are 11.47% and 2.38%, respectively. The simulation results agree well with the experimental results. It proves that the finite element model is reliable. The Mises stress nephogram, the contact area between soil and press roller and the displacement nephogram are adopted to compare the simulation results of the two press rollers. The simulation results show that contact stress between the bionic roller and soil is significantly higher than that of conventional roller, which is beneficial for crushing clods. The contact area between bionic roller and soil is 79.05% lower than that of conventional roller, which is helpful to reduce soil adhesion. Also this result is proved by the soil bin test results, which shows that the bionic roller can reduce 52.78% of soil adhesion. Adjacent ridge structures of the bionic roller can properly constrain the flow of soil, thus avoiding the hilling phenomenon in compaction process. The two press rollers show little variance in vertical displacement and have similar compaction resistance, but bionic roller is more conducive to compact topsoil. The soil displacement in the direction of roller width caused by the conventional roller is greater than that of the bionic roller, which makes the conventional roller consume more power. The soil bin test results show that the bionic roller lowers the resistance by 28.66% as compared with conventional roller.