This paper focuses on the problem of preference satisfaction in service planning of a group of service robots, especially on how to realize a feedback architecture for enhance service quality. Preferences from service users and service robots are set as arguments in an argumentation framework. The argumentation process leads to a mixed set of preferences and constraints without inner conflicts, excluding the preferences that cause conflicts. The argumentation result are utilized to generate service plan via a projection-like algorithm, and also sent back to users and robots, who may change there habit on raising preference so as to avoid conflicts in the next planning. Preference considered including static preference and temporal preference. Case study on office service is proposed to show the effectiveness and characteristics of the proposed framework.