This paper deals with a switching control method for underactuated manipulators. This method selects a partially-stable controller from a number of partially-stable controllers which are designed fuzzy reasoning. We design switching rules on an error energy plane, in which boundary curves to split an energy plane into utilized partially-stable controllers describe fuzzy rules. In this study, we consider a tree link underactuated planar manipulator which has a passive joint and two active ones. In this paper, we present an evaluation of installation positive joint through some simulations. Design parameters of boundary curves and partially-stable controllers are obtained by genetic algorithms (GAs).