To solve the autonomous urban driving problem presented by the 2007 Defense Advanced Research Project Agency Urban Challenge, team VictorTango developed a tri-level architecture with a deliberative-reactive-deliberative structure. The VictorTango architecture emphasizes a robust, reusable, and modular design, using hardware independent virtual actuators and virtual sensors. Details of the Urban Challenge implementation are provided, highlighting the functionalities and interactions of different modules within the architecture. Situational examples from the Urban Challenge problem illustrate the performance of the architecture in real-world situations, and communications latencies with their effect on decision making are analyzed. Finally, recommendations for future refinement of the architecture are presented. The VictorTango architecture ultimately proved capable of completing the Urban Challenge problem. Furthermore, the modularity, hardware independence, and robustness of the architecture have enabled it to be applied to other platforms and other problems in the Unmanned Systems field.