Abstract: Hierarchical strategy can be used in solving the complicated robot plan-generating problem.This paperproposes a new hierarchical planning algorithm.It splits the primitive Problem into several subproblems,then solves the action sequences from the top to the bottom according to the action accuratization rule till theprimitive one is finally solved.As the application of nonlinear technique in different strategy makes it feasibleto solve the subproblems alternatively,it settles the interacting goal problem.With the formulation of an evaluation function and forward & backward pointers in the process of plan generation,it also finds the way toeliminate the redundance and contradictions in hierarchical planning,and therefore greatly improved the efficiency of problem solving.The algorithm has been applied to NARP robot plan-generating system,which uses Common Lisp Language in programming and has been implemented on SUN computer.