Road collapse poses a significant threat to urban infrastructure, especially in geohazard-prone areas and critical facilities, involving complex interactions among discontinuous solids, pore fluids, and structural components. Real-time monitoring of subsurface and surface deformations, combined with numerical simulations, enables proactive geohazard warnings and is therefore crucial for improving infrastructure resilience. The Material Point Method (MPM) has emerged as a promising approach for capturing ground subsidence processes and elucidating the underlying soil-water interaction mechanisms. In this study, a two-phase double-point MPM program was developed to accurately simulate the evolution of cavity formation and subsequent ground surface collapse. The method employs two distinct sets of material points to represent the soil (solid) and water (fluid) phases, respectively, while their interactions are mediated through a background grid using high-order B-spline basis functions. The effectiveness of the modified MPM scheme was first validated via a benchmark slope stability analysis. Subsequently, a series of laboratory model experiments were conducted to further assess its predictive capability. The MPM simulation results show good agreement with experimental observations, confirming the feasibility and accuracy of the proposed method in handling large-deformation cavity formation and ground subsidence under fully saturated conditions.