[en] An approach to fuel management optimization using differential dynamic programming is presented. At each refueling period a decision must be made which determines the vector state of the fuel batch, defined as the average isotopic composition of the fuel for a given fuel region. The recycling of this fuel batch results in a multi-stage decision process with the levelized fuel cycle cost as the objective to be minimized. The vector state of the fuel after the end of each stage may be found using a state transition matrix determined from a depletion routine which may accommodate any degree of complexity in the reactor physics computations. The application of differential dynamic programming requires that the levelized fuel cycle cost function be expanded in terms of a truncated Taylor series about a set of state vectors. The coefficients of the series are determined by a nominal trajectory of state vectors over a given number of refueling stages. A search is conducted about this nominal trajectory until a more optimal trajectory is found which in turn becomes the nominal trajectory for the next iteration. Applications to batch fuel refueling, partial fuel refueling and fuel shuffling, and a recycle program between a LWR and LMFBR are presented. (U.S.)