INTELLIGENT CONTROL OF MATERIAL HANDLING SYSTEMS - Environmentally Conscious Materials Handling - page 76

77

9 Agent Communication

2. The plan layer is on top of the behavior layer. It handles all the

goal-directed and proactive planning of the agent.
3. A cooperation layer describes the collaboration and interactions with other
agents.
A common problem with hybrid architectures such as InteRRaP is that
there is no clear semantics or methodology for programming such architectures
(Wooldridge 1999), so it can be hard for the developer to design a coherent
agent behavior.

9 AGENT COMMUNICATION

Interagent communication is a key requirement for an agent to fulfill its social
characteristics, so a long list of interaction schemes and communication protocols
have been proposed and implemented in the agent community. However, with
respect to manufacturing and material handling systems, particular coordination,
negotiation, and hybrid mechanisms are dominating (Bussmann et al. 2004).
Apparently, coordination is a form of communication — not only for
agents — that will be most noticeable if it does not work properly. Perfectly
aligned conveyors, input, and output facilities of material handling systems
require a high degree of coordination between the control elements, which are
obtained through an often long and tedious alignment process at installation
time. For flexible manufacturing systems, the conditions for coordination are
constantly subject to changes. Therefore, coordinating agent activities is of
highest priority in intelligent manufacturing and material handling systems.
With a system composed of flexible cells and connected through, for example,
conveyors, coordination can be defined as process of managing dependencies
between activities (Malone and Crowston 1994). Coordination is not a trivial
thing to achieve in agent-based control systems. Jennings has emphasized
three common characteristic of agent systems that lead to dependencies and
complicate coordination (Jennings 1996):
• Actions of agents might interfere . Two agents might fight for the same
resource in order to complete their tasks.
• Global constraints might have to be satisfied . It could be suitable to dis-
tribute the load on the entire system, but there will still be an overall
deadline of an order that an item must satisfy.
• Individual agents cannot satisfy their own or system goals by itself .The
core idea of a flexible production cell is that an item has to process several
work stations for it to be completed.
Malone and Crawston (1994) further simplified the interdependencies relevant
for distributed agents into three types of dependencies, as shown in Figure 3.7.
A flow dependency arises if one task in the process produces or generates a
resource that is required by another task. This is the most common dependency