Application Design

MedPAge Technical Infrastructure

As shown in the figure above, the MedPAge multi-agent system consists of different layers that are based on two fundamental building blocks. On the one hand an agent infrastructure is needed that provides functionalities for agent lifecycle management, communication means and services for agent registration and search. For this purpose the FIPA-compliant JADE multi-agent platform is used. On top of the basic agent functionalities a rational agent layer is placed. It allows the usage of higher-level abstract concepts for agent programming following the belief, desire, intention theory. With these mental attitudes it is possible to specify the agent's internal behaviour in a goal-oriented way that resembles the way humans think and act. For this purpose the Jadex extension for JADE is utilized.

On the other hand the MedPAge system needs a persistent data storage for saving static hospital model information as well as dynamic reservation and treatment data. For reasons of legacy hospital information system integration and robustness a relational database was chosen. The concrete choice is MySQL, mainly because it comes along with an outstanding web database administration tool called PHPMyAdmin and therefore simplifies the installation and testing of the MedPAge system. To transfer the relational data their object-oriented counterparts transparently, an object-realtional mapping framework is utilized. Currently the Cayenne framework is used.

On top of these two fundamental blocks the hospital simulation layer is arranged. It enables the MedPAge system to be tested efficiently in a controlled environment. Heart of the simulation layer is the time synchronization component, which is responsible for managing and advancing a global simulation clock. This component is realized as a specialized agent service based on event-based process simulation foundations (1). The hospital model contains information about the static hospital structure, like available resources including personnel data and dynamic patient-related data like appointments or results of treatments. The Agent.Hospital Interface represents the access point of the MedPAge system to the outer world. This outer world is embodied by Agent.Hospital, which is a test-scenario consisting of several interconnected multi-agent systems (Kirn et al 2003). With the help of the gateway agent concept and a common Agent.Hospital ontology including service descriptions of all service providers it is possible to use MedPAge services from the outside as well as using other services within Agent.Hospital from the MedPAge system.

The hospital scheduling process is mainly composed of two activities. A scheduling mechanism that runs between patient and resource agents and delivers a schedule and the execution of treatments that is directly related to the actual (simulation) time. The second task includes, that the patients with a due appointment are called to the appropriate resources just in time. This is generically done by a so called Call-Patient-Module which resides in the hospital simulation layer. The scheduling strategies are therefore easily exchangeable and compose the coordination layer. Of main interest for the MedPAge system is a strategy called MedPaCo which is based on a decentralized negotiation approach utilizing worth functions (5).

• Stefan Kirn, Christian Heine, Rainer Herrler, Karl-Heinz Krempels
  "Agent.Hospital - agent-based open framework for clinical applications"
  In Proceedings of 1st International Workshop on Agent-based Computing for Enterprise Collaboration, WETICE 2003