The Discrete Unified Gas Kinetic Scheme (DUGKS) has become an effective method for simulating multiscale particle transport problems. Unlike conventional direct interpolation schemes, DUGKS reconstructs the distribution function at cell interfaces by employing a trapezoidal integration of the kinetic model equation along characteristic lines in both time and space, thereby simultaneously accounting for free transport and collision effects of particles. A key numerical feature of DUGKS is that the interface distribution function is reconstructed from the interpolated values of multiple neighboring cells. However, such neighbor-dependent reconstruction may reduce accuracy and robustness for strongly heterogeneous multiscale transport problems, especially on coarse meshes, and introduces additional burdens for parallel implementation. To address these issues, this paper proposes a Compact Discrete Unified Gas Kinetic Scheme (CDUGKS) together with a parallel solution framework. The CDUGKS employs a general compact reconstruction strategy that determines the interface distribution function using only local information within a single cell. Numerical results from various particle transport benchmarks demonstrate the superior accuracy of the CDUGKS, particularly for problems involving steep gradients and strong material discontinuities. Moreover, the spatial compactness of CDUGKS makes it especially suitable for large-scale parallel computations.

Discrete unified gas kinetic scheme,Multiscale particle transport,Large-scale parallelism,Compact reconstruction strategy