Loose asphalt concrete is a commonly encountered defect in road surface layers, often leading to premature issues such as pavement cracking and potholes, thereby compromising the overall performance and lifespan of the road. Leveraging the statistical characteristics of asphalt concrete's multiphase, discrete, and random distribution, we employed a quantitative constraint multiphase discrete random medium modeling approach to develop models of asphalt concrete loosening with varying porosity rates. Additionally, we conducted GPR (Ground Penetrating Radar) forward modeling to investigate the GPR wavefield characteristics and intuitive diagnostic techniques associated with loosening. Our research findings reveal that, in comparison to traditional layered uniform medium models, the multiphase discrete random medium model offers a more precise portrayal of the actual state of asphalt concrete loosening. Furthermore, its numerical simulation outcomes align more closely with measured radar data. As the degree of asphalt concrete loosening increases, the inhomogeneity of the medium becomes more evident, resulting in a stronger amplitude of the corresponding GPR wave, a more chaotic waveform, more intense variations in regional amplitude curves, and more prominent diffraction waves on both sides. By analyzing the GPR waveform chaos, amplitude intensity, and intensity of change, we can more intuitively and accurately identify loose areas in asphalt concrete, qualitatively assess the degree of loosening, and provide a solid foundation for targeted treatment and repair measures.