Themostefficient planning algorithms recently developed are mainly based on Graphplansystem or on satisfiability approach. In this paper wepresent a new approach to plan generation based on planning graph analysis, whichcan be considered as a bridge between the two planning approaches. Themethodexploits the propagation of planning axioms and constraints in order to makedeductions on the planning graph and therefore to prune the search space. The consequences of decisions madeduring search have backwardand furward impact on the planning graph. In contrast with Graphplan based backward algorithms, our approach allows to search the planning graph without committingto any specific direction. Theexperimental results obtained with DPPlan, a planner implementingthe presented propagation approach by systematic search, are encouraging even if comparedwith approaches based on SAT. DPPianhas not the huge memoryrequirements as SAT solvers and keepsa strong connectionwith the planning problemallow...