Intelligent agents in logistics.
Report 2000.LT.5302, Transport Technology, Logistic Engineering.
Conventional logistic systems are large centrally managed control systems
based on top-down architecture. Fundamental changes in economic, social
and legislative factors have made market demands for logistic systems
become more complex. New logistic systems should be adaptable, changeable,
scaleable and network-oriented. The objective of this paper is to find out
if and how intelligent agents can be used in new logistic systems.
The concept of agents comes from Minsky who tried to explain how the mind
works, using a model made of many small processes, called agents. Agents
are software components that can perceive, reason, act, communicate with
other agents, have a social autonomy, can modify their own behavior and
which can be heterogeneous.
Agents are best developed in multiagent systems (MASs), but most agent-based
systems currently consist of a single agent. Neither systematic methodology to
design MAS-applications nor a widely available MAS-toolkit does yet exist.
Communication is an important issue in the development of MASs. Two main
Agent Communication Languages (ACLs) do exist now: KQML and Arcol. These
ACLs are focussed on pragmatics (the way symbols in the ACL are
interpreted and used) instead of on commitments. Focussing on commitments
is a promising solution.
In this report three different examples of application of intelligent
agents in logistics in different fields are discussed: agents in job flow
scheduling, in simulation software and in urban traffic control.
Although theoretically it is possible to build an agent-based logistic
system that meets all requirements for new logistic systems, in practise
this still is difficult. The application with the best-structured
architecture has the least problems with meeting the requirements.
Reports on Logistic Engineering (in Dutch)
, TU Delft