Vehicle tracking can be done automatically based on data from a distributed sensor network. The determination of vehicle behavior must currently be done by humans. Behaviors of interest include searching, attacking and retreating. The purpose of this paper is to show an approach for the automatic interpretation of vehicle behaviors based on data from distributed sensor networks. The continuous dynamics of the sensor network are converted into symbolic dynamics by dividing its phase space into hypercubes and associating a symbol with each region. When the phase-space trajectory enters a region, its corresponding symbol is emitted into a symbol stream. Substrings from the stream are interpreted as a formal language defining the behavior of the vehicle. The formal language from the sensor network is compared to the languages associated with known behaviors of interest. Techniques for performing quantitative comparisons between formal languages are presented. The abstraction process is shown to be powerful enough to distinguish two simple behaviors of a robot based on data from a pressure sensitive floor.