Any realistic plan specification must take into account temporal and resourse restrictions for actions. The classical approach fo executing plans with restrictions works in two alternating phases. During the first phase, the set of actions that are ready to be executed is determined. In the second phase, temporal and resource restrictions are taken into account to generate a viable scheduling for the ready actions. This separation into two phases may lead to inefficiencies, when inconsistencies in the second phase force backtracking to the first phase. This dissertation first proposes a plan model that incorporates a rich language to specify temporal restrictions. Then, it introduces a plan execution algorithm that integrates the two phases mentioned above, thereby reducing as mush as possible the need for backtracking.