Abstract: To elucidate the rules of particle crushing and loss of gravel soil subgrade filler during compaction and penetration, we performed compaction and permeability tests of gravel soil samples with different stone and moisture contents; moreover, crushing condition and fractal analysis were also performed on the test results. The results reveal that the best compaction effect is achieved when the stone content is 60% for gravel soil samples with hard rocks and fine particles as sandy soil and optimal moisture content is 8%. With the exception of the 70% stone content, the permeability coefficients of each ratio are similar, and the permeability coefficients all range from 10⁻² to 10⁻³, indicating good permeability. The particle size distributions before and after compaction and crushing exhibit good fractal characteristics. The fractal dimension D is positively correlated with the dry density. Furthermore, the optimal fractal dimension D for crushing is 2.23–2.25 and 2.43–2.45 for coarse and fine particles, respectively. Relationships between the fractal dimension D and stone and moisture contents are established. Based on the relationship and optimal fractal dimension interval, the fractal dimension D can be predicted, and the compaction effect can be further estimated. There is also a good linear relationship between the crushing ratio B_g and fractal dimension D. When the stone content is no more than 60%, D of the coarse and fine particles increases with increasing B_g. The relationship between the fractal dimension D and penetration erosion shows that the smaller the difference between the fractal dimension before and after penetration, the smaller the effect of penetration on the erosion of gravel soil. The permeation erosion effect on D is highly affected by the stone content. When the stone content is 60%, the difference in fractal dimension decreases first and then increases as the moisture content increases, which is consistent with the changing trend of the permeability coefficient. Moreover, stone and moisture contents are two important factors that affect particle crushing and loss. However, stone content has a more significant impact on particle crushing and loss than moisture content. The fractal dimension D has a good correlation with the compaction and permeability test results of each proportion sample, which can better reflect the compaction and permeability characteristics of gravel soil samples. Thus, the results of this study can further reveal the compaction mechanism of gravel soil subgrade and provide a reference for subgrade construction.