Systems and methods for controlling automotive powertrains using a distributed control architecture are disclosed. A distributed control system may include a supervisory control unit for controlling one or more powertrain subsystems, and one or more subsystem control units in communication with the supervisory control unit. The supervisory control unit can be configured to execute a central optimization algorithm that computes variables from across multiple powertrain subsystems, and then outputs a number of globally approximated command values to each associated subsystem control unit. In some embodiments, the central optimization algorithm can be configured to solve a global cost function or optimization routine. One or more of the subsystem control units can be configured to execute a lower-level algorithm or routine, which can comprise a higher-fidelity model than that used by the central optimization algorithm.