Agencies of Agents for Logistic Applications
Sonali Banerjee · Studies in fuzziness and soft computing · 2002
This paper presents an overview of current trends in Electronic Business (E-business), and discusses how an enterprise can use the Electronic Marketspace to its strategic advantage. In this paper, we define an agency to be a multi-agent system created by integrating agents selected from a library of reusable agents that have formed a federation. A federation of agents comprises of a set of registered agents, which are themselves complete knowledge-based systems. The set of agents may be heterogeneous with respect to long-term knowledge, solution-evaluation criteria, or goals, as well as languages, algorithms, hardware requirements, etc. An agency-based framework is presented for E-business in the domain of logistics and supply systems. To support the integration of heterogeneous and reusable agents into functional agent sets, a framework for the coalition of cooperating agents is needed. The role of cooperative information agents is discussed within the context of E-business. Conflict is an integral part of problem solving in multi-agent systems and is often the focal point of interaction among agents. Rules about agent interaction can be used to mediate, mitigate or avoid conflict and as a control structure for agent activity to ensure global coherence of the agent set. This can be achieved by an executive agent that has an overall perspective of the other agents in the federation. A facilitator, providing query and result refinement, together with, decision making tools that can be customized to take advantage of specific agents and agent-set characteristics, while maintaining a global view of the problem that needs to be solved can be integrated with the adoption of the agent architecture discussed here. The effectiveness of the facilitator as a coordination mechanism is in the specification of the rules that guide the task or service agents in problem solving and the meta-rules which instantiates appropriate knowledge sources. Several scenarios, with the prototype facilitator has been considered as proof of this concept. Such logistic systems enjoy wide acceptance in commercial and military application domains, as witnessed by the phenomenal growth in Internet stocks for companies such as Amazon.com. We show how agent-based systems provide scalable, configurable and evolutionary solutions for such applications. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.