GCC based agents collaboration mediation
Wei Liu, Jing Wang, Deng Chen · 2018
Autonomous agents collaboration at run-time is a fundamental part of multi-agent oriented adaptive collaboration systems (ACSs) when an agent cannot accomplish a task by itself under changing environment. Such collaborations are usually regulated by commitment protocols, which are typically defined at design-time. However, agent commitments among the participants to meld collaborative patterns typically defined in static environments, which may not fit the need of ACSs. To address this, we propose a Goal-Capability-Commitment (GCC) based mediation for agents collaboration at run-time that combines invalid diagnoser, capability reconciliation and commitment compensation. We elaborate on each of GCC elements, focusing in particular on the feasibility afforded by (1) the ability to diagnose the failures of capability and commitment leaded by the changing environments, and (2) the ability of mediation to select appropriate capability or generate new commitment in order to satisfy the invalid goals. We represent the GCC model in a hospital medical waste AGV transportation case study and illustrate our approach through experiments.