In response to high costs, monotonous scenarios, and insufficient situation awareness during verification process of unmanned driving technology in the airfield area, a semi-physical verification platform was developed. The platform was divided into three parts: physical layer, model layer, and simulation layer. In the physical layer, a local area network was established with onboard sensing equipment, to generate digital-twin trajectory of realistic vehicles. In the model layer, to meet the needs of customized design for unmanned vehicles, a data exchange method between flight simulator and unmanned driving vehicle simulator was provided. Hence, precise reproduction of vehicular or aeronautical kinematics models could be achieved. Dynamic conflict protection zone for movement target was designed, to predict conflicts and react for them. In the simulation layer, simulation scripts were established to create global environment for verification. Test results from the platform indicate that stopping distance of single unmanned vehicle during initiative avoidance process is 6.174 m; velocity deviation and lateral deviation of the digital-twin trajectories are within the standard deviation ranges of [−0.51 m/s, 0.65 m/s] and [−0.75 m, + 0.75 m] respectively. The results show reproduction ability of operational situation, meeting the needs of economical and multi-scenario verification.