Proaktive autonome Navigation für mobile Agenten: ein Schritt in Richtung mobiler Agentensysteme der nächsten Generation

Logical or virtual networks based on mobile agents exhibit a high level of internal dynamics: Agent servers and services are added, deleted, or modified, connection quality changes over time, etc. In addition, the size of these networks steadily grows, as practical experience shows, and adds a quantitative factor that makes scalable solutions and structures mandatory. These aspects cannot be handled by the current state-of-the-art in mobile agent technology. Today’s systems expect a mobile agent’s user and/or programmer to know the current structure of the underlying network and to keep this knowledge up-to-date all the time. The agent is programmed with a fixed itinerary (route) through the network that is based on the current knowledge the programmer has to offer. This leads to problems with the dynamical and quantitative aspects of the network and may cause incomplete or even incorrect routing of the agent on its itinerary. This thesis presents the generic framework called ProNav that offers functions to manage both mentioned network aspects. ProNav is designed as an extension of mobile agent systems that provides independent routing on the application level, i. e. offers autonomous and proactive navigation that takes into account the dynamics and size of real world networks. From agent’s perspective, ProNav is just another (infrastructure-)service provided on each agent server. Its most important feature is to locate services and information in the network and to offer this type of data to any mobile agent currently planning its itinerary. This is achieved by integrating the data that is locally acquired by each agent server into a map that enables each agent to recognise and analyse its virtual environment. For a mobile agent the opportunity for a proactive and autonomous planning of its itinerary is offered by ProNav, even if the network is highly heterogeneous and dynamic. The agent’s programmer and user do not have to do the itinerary planning for the agent. The agent is enabled to fulfill this task itself and independently of its owners. ProNav also provides enough information and flexibility to abandon the notion of a fixed route through the network and allows regular updates and changes in the itinerary during its execution. This helps to react immediately and in an autonomous fashion to changes in the environment. From an architectural perspective, ProNav extends any mobile agent system by working as an intermediate layer in-between the actual agent system and the application layer that is formed by specialised mobile agents and their user and application interfaces. ProNav is divided into three major components: the map module, the route planner, and the migration optimizer. In principle each component may be used independently by any mobile agent. However, only by integrating their services a mobile agent will achieve full autonomy and pro-activity for the itinerary planning task. Each map module keeps its data up-to-date by using the local resources of its agent server. These local maps are integrated by means of the route planner module into a more global view that allows for incremental itinerary planning for each specific mobile agent. The migration optimizer can be utilized to improve further the technological implementation of an itinerary, e. g. to optimise time or cost factors on the route. This module interacts directly with the underlying mobile agent system’s migration component. ProNav may be used with any mobile agent system that offers a compatible infrastructure organization. The current prototype, also used during the evaluation phase of the project, relies on the mobile agent system TRACY that was developed at the Friedrich-SchillerUniversity Jena.