A combined model for time-dependent trip distribution and traffic assignment is proposed. The model assumes known time-dependent departure rates from origins and overall arrival rates at destinations, and it then seeks to estimate the origin-destination matrix according to the observed entropy value and to minimize the total system travel time. A solution algorithm is proposed based on the Dantzig-Wolfe decomposition principle and a Lagrangian relaxation approach. Computational results on a test network are also presented and discussed.