We propose an implemented architecture that integrates possession of environmental knowledge into a reactive, behavior-based autonomous robot. As an alternative to hybrid combinations of behavior- and model-based systems, we present a fully behavior-based system that takes reactive, deliberate behavior. To get knowledge, a robot makes a the World Image, gradually constructed through experience as higher animals with self-organization do, without symbolic representation in associative memory. Since our World Image changes in direct relation to behavior, a robot navigates adaptively and reactively to its destination following a path plan based on the World Image of a dynamic environment. The path plan a robot makes is not a sequence of routes as in conventional planning, but flexible as in reactive planning. If a robot detects a failure, it reaches its destination without replanning or error recovery. Our approach makes practical a behavior-based robot already adapted to the real world.