Robot systems are highly complicated systems composed of various hardware and software components. To develop robot systems more efficiently, there have been many studies about the robot development platforms. However, such platforms consider more on the efficiency and productivity at the development level for developers rather than the convenience at the service level for users. Considering convenient design of robot systems at user level, the robot system should be reconfigurable with hardware and software components by the users. To deal with such problem, this thesis proposes a novel methodology for the development of robots enacted through user-level agent-based modularization (DREAM) and provides the design goals for the DREAM methodology, which are `hardware and software components modularization,` `cooperative communication between components,` `components management at service level,` and `two levels of abstraction for a behavior translation.` To realize the design goals, the agents manager is implemented as a core program based on the concepts of agent and message. As a result, the implemented robot system based on the DREAM methodology enables users to efficiently and comfortably build or reconfigure their robots for their own usage. Moreover, for the independent development and replacement of agents, this thesis proposes a script-based behavior translation method for the DREAM methodology. A behavior translation agent is implemented for converting a behavior, selected by an application agent, into a sequence of command data for an actuator agent using behavior scripts. As a result, application agents can work flawlessly even if hardware components are replaced and vice versa. The effectiveness and applicability of the proposed DREAM methodology are demonstrated through illustrative experiments and the related applications in a real world environment.