A METHOD of decomposing a linear dynamic programming problem is discussed, which amounts to projection onto the space of state variable s The resulting problem of piecewise linear programming can be reduced to a finite set of linear programming problems by constructing a division of the set of admissible state values into domains of linearity of the functional. Two algorithms are described, for the general and non-degenerate cases, convergent to the exact solution of the initial problem after a finite number of steps.