A method for generating training data. The method may include executing a simulation process. The simulation process may include traversing one or more virtual sensors over a virtual road surface defining a plurality of virtual anomalies that are each sensible by the one or more virtual sensors. During the traversing, each of the one or more virtual sensors may be moved with respect to the virtual road surface as dictated by a vehicle-motion model modeling motion of a vehicle driving on the virtual road surface while carrying the one or more virtual sensors. Virtual sensor data characterizing the virtual road surface may be recorded. The virtual sensor data may correspond to what a real sensor would have output had it sensed the road surface in the real world.