An important reason for the popularity of the behavior-based paradigm in autonomous robotics is the possibility to design complex robot behaviors in an incremental way. We propose a fuzzy hierarchical behavior-based architecture, in which rules and meta-rules are used in a uniform way at all levels of the control hierarchy. This architecture has been successfully used in a number of robots performing autonomous navigation tasks. In this paper, we show the use of hierarchical fuzzy behaviors to implement a set of navigation and ball control behaviors for a Sony four-legged robot operating in the RoboCup domain. We also show that the logical structure of the rules and the hierarchical decomposition simplify the design of very complex behaviors, like the ``GoalKeeper'' behavior.