This paper presents a middleware architecture for personal robots when applied to various environments. This architecture allows a user's personal robot to consistently integrate environment-oriented applications to the robot's personality in its own application. These applications share sensors and actuators, and generate consistent actions in sequence. However, integration of two different kinds of independently developed applications causes a lack of its continuity and a slow response in human-robot communication. To solve these problems, we have designed a new middleware f o r personal robots, called P R I M A (Personal Robots' Intermediating Mediator Adaption to environment) which can schedule the outputs of these applications to keep open the human-robot communication channel. W e call the integration of these applications a personalization of dynamically loaded service programs. In addition, our experiment demonstrates the effects of P R I M A o n human-robot communication.