A high-fidelity physics-based approach for predicting vehicle mobility over large terrain maps is presented. The novelties of this paper are: (i) modeling approach based on seamless integration of multibody dynamics and the discrete element method (DEM) into one solver, and (ii) an HPC-based design-of-Experiments (DOE) approach to predict the off-road soft soil mobility of ground vehicles on large-scale terrain maps. A high-fidelity multibody dynamics model of a typical 4x4 military vehicle is used which includes models of the various vehicle systems such as chassis, wheels/tires, suspension, steering, and power train. A penalty technique is used to impose joint and contact constraints. A general cohesive soil material DEM model is used which includes the effects of soil cohesion, elasticity, plasticity/compressibility, damping, friction, and viscosity. To manage problem size, a novel moving soil patch technique is used in which DEM particles which are far behind the vehicle are continuously eliminated and then reemitted in front of the vehicle as new particles and then leveled and compacted. The governing equations of motion of both the vehicle and the soil particles are solved along with joint and contact constraints using an explicit timeintegration procedure. The DEM inter-particle cohesion and friction forces are calibrated to the cone index using a simulation of a cone penetrometer experiment. The DOE approach is demonstrated by predicting the speed-made-good distribution for a typical military vehicle on 22 km × 22 km terrain map. Two terrain parameters are considered in the DOE, namely, terrain positive slope and soil strength. This is the first time such mobility map is generated using highfidelity physics-based simulations.