INTELLIGENT CONTROL OF MATERIAL HANDLING SYSTEMS - Environmentally Conscious Materials Handling - page 69

production environment, and mass customization overtook the efficiency of CIM.
Instead, the responsibility in CIM systems was distributed to autonomous, intel-
ligent, and collaborating components, which led to the holonic manufacturing
approach under the IMS program already mentioned.
The shortcomings of the holonic manufacturing gradually prepared the
agent-based approach as the most promising software technology for intelligent
control. Whereas holonic systems have the focus on all the mechatronic
components of an IMS, an agent model of the system also incorporates the
planning, scheduling, and interoperability among the agents.
Multiagent-based systems (MAS) are still a relatively new paradigm in com-
puter science, which can suit many other purposes than control of logistics and
manufacturing systems. MAS facilitate an optimization of the decision process
and add an extra level of robustness and stability to complex, heterogeneous
systems. Agents are able to interact in dynamic, open, and unpredictable envi-
ronments with many actors — here called agents , who cooperate to solve specific
tasks or achieve design goals.
Usually, an agent-based manufacturing and material handling system is mod-
eled with agents in all the decision points of the systems, such as assembly
stations, employees, cranes, AGVs, robotic cells, PLCs, and so on. These agents
will be able to act autonomously by observing their own local neighborhood and
communicate with other agents. An agent will also act and change its actions
according to the current status of its environment, so it can achieve its design
goals as best as possible. This makes the system robust to local unpredictable
events, as they will only be perceived by the relevant agents, who will adjust
their actions, and the effect will propagate throughout the system through the
interagent communication.

5 AGENT TECHNOLOGY

Agent-based, or multiagent, systems had emerged from artificial intelligence long
before they were considered for control in manufacturing processes. The research
area of artificial intelligence was born in the late 1950s and was focused on both
understanding the human reasoning process and developing methods and tools
to built intelligent systems. In the first decade, expert systems were the primary
base for research in artificial intelligent systems. The decision process of the
systems was usually modeled as condition-action rules that were triggered by
events from the environment or changes in an internal world model.
Pattern matching and understanding natural languages were hot topics of the
time for such type of systems. It was natural to have different knowledge sources
that work on different aspect of the problem, which again led to the notion of
distributed artificial intelligence (DAI). Different expert systems were used to
partially process recorded data in the HEARSAY speech understanding system
(Erman and Lesser 1975). Erman and Lesser used a blackboard architecture to