A middleware support for agent-based application mobility in pervasive environments

Abstract

Application mobility is an efficient way to mask uneven conditioning and reduce users' distractions in pervasive environments. However, since mobility brings more dynamism and uncertainty, it also raises new research issues in developing pervasive applications, including underlying application models, adaptive resource rebinding mechanisms, synchronization and fault tolerance techniques, etc. In this paper, we approach these problems from the middleware perspective. Inspired by software agent's inherent capability of autonomy and mobility, we investigate its potential use in application mobility and propose an agent-based architecture called MDAgent. Three salient features are emphasized: 1) Reduced mobility overhead. Flexible bindings of application components avoid migrating whole application. 2) Simplified mobility management. Mobile agent takes over the responsibility of mobility and synchronization, so user intervention is reduced. 3) Enhanced customizability and adaptability. Context information can be updated dynamically, and ontology-based reasoning ability embedded in autonomous agents can direct the application to adapt to the changes accordingly. On top of MDAgent, we have developed several applications, and evaluated the performance.