Importance as well as difficulty of pavement management planning at the project level lies in pavement rehabilitation timing and selection of rehabilitation methods in order to minimize maintenance and rehabilitation costs during pavement life time. Timing and rehabilitation methods make planning difficult and less reliable due to the actual life of pavement differing from its design life and tending to vary widely for various reasons. This paper proposes an optimal method for pavement management planning based on probabilistic dynamic programming (PDP), which can include uncertainties in a statistical sense. Efficiency and reliability of PDP are demonstrated through example problems.