Multi-agent systems provide a possible approach for solving distributed problems. Unlike conventional distributed problem solving strategies multi-agent systems can handle distinct and maybe competing goals in the same system. The IEEE Foundation for Intelligent Physical Agents referred to as FIPA defines standards for multi-agent systems by specifying system components at functional and not implementation level.
Its aim is to promote commercially usable agent-based technologies. Additionally, FIPA specifies so called interaction protocols serving agent coordination. FIPA-compliant agent platforms do not directly offer multi-agent planning facilities. This master's thesis presents a new approach for integrating multi-agent planning and scheduling in a FIPA-compliant multi-agent infrastructure.
Framework for Task Analysis, Environment Modeling and Simulation referred to as TÆMS provides a formal domain-independent framework to represent qualitative and quantitative aspects of coordination as well as a model for solving multi-agent planning and scheduling problems under soft realtime conditions. Vienna Criteria-Driven Scheduler referred to as VIE-CDS is a Design-to-Criteria scheduler working on TÆMS structures. So called scheduling criteria allow to determine preferences for schedules concerning quality, cost and duration of schedule execution.
In this master's thesis, the process of scheduled action execution is defined by five phases using only FIPA standardised interaction protocols. This approach guarantees applicability to FIPA compliant multi-agent systems.
Based on the new XTAEMS encoding of TÆMS structures, an example implementation using the Java Agent Development Framework referred to as JADE, being a representative of FIPA-compliant multi-agent infrastructures, is provided.