Abstract: Objectives: To address the challenges of vehicle trafficability assessment in roadless environments, this study proposes a remote sensing inversion method for evaluating vehicle support mobility on soft terrain, validated in Zhangzhou and Chuzhou test areas. Methods: A gradient boosting tree model was developed to invert shallow soil moisture using remote sensing data, supported by satellite-ground synchronous experiments. A knowledge distillation model was constructed to classify soil USCS types, verified through in situ measurements. Soil mechanical parameters and vehicle dynamics models were integrated to predict sinkage and speed for tracked/wheeled vehicles on digital road networks. Results: The soil moisture inversion achieved RMSE values of 6.3% and 4.9%, demonstrating high accuracy. The USCS classification model showed 75.0% consistency with field data. Vehicle sinkage predictions exhibited errors below 2 cm, and speed estimation accuracy exceeded 90% in practical validations. Conclusions: The proposed full remote sensing inversion method enables efficient calculation of maximum vehicle speeds, supporting navigation path planning and vehicle guidance in unstructured environments. This approach advances the application of remote sensing technology in mobility assessment and analysis.