wilamowski-b-m-irwin-j-d-industrial-communication-systems-2011

Industrial Agent Technology 16-9
An order holon has simpler logic but is more complex in hierarchy, meaning that it can divide a
given order into suborders that are controlled by a parent holon of a higher layer. It mainly represents
the customer request, where all requirements on the product are defined. It has an advantage in mass
production because an order holon is not attached to a single work piece. An order holon does neither
deal with scheduling tasks nor resource allocations because it simply does not have knowledge
about the physical layout of the plant and the product specification. The latter, instead, is managed
by a separate entity—the product holon. It is nothing more than a “recipe” for a particular type of
product. A disadvantage of this solution is the lack of customization of products—a major reason for
distributed automation systems.
Similar to the other aspects of general patterns there is always a balance between optimization
and flexibility and hierarchy of orders and management entities related to them. While PROSA organizes
orders in a strong hierarchy and PABADIS minimizes interdependencies between the orders,
PABADIS’PROMISE provides what can be called an “implicit hierarchy.” There is no strict decision
making like in PROSA. The organization is implemented via the structure of the production order that
reflects interdependencies between different production steps (called process segments) via so-called
node operators. The mechanism of order decomposition developed in PABADIS’PROMISE allows
assigning autonomously acting OAs to each process segment. It is similar to the PABADIS approach
with the difference that higher level OAs can influence the agents responsible for subtasks execution.
For instance, the OA that is responsible for the assembly of a car needs an engine to be produced before
performing its tasks. Therefore, it assigns the deadlines for the engine agent, so the engine is delivered
in time.
Another issue is the behavior of agents involved in the production. PROSA with its hierarchy does
not need to pay attention to this topic, but in more distributed systems such as PABADIS it has higher
importance because it can increase efficiency and performance. Product agents in PABADIS are selfish
in nature. As a result, the system provides lower optimization of the order execution compared to
PROSA, but much more flexibility. PABADIS’PROMISE aims to combine both benefits by introducing
benevolent behavior of agents; that is, while agents rely on a certain set of rules that benefit their local
goals (e.g., finishing assigned process segments on time) they may, nevertheless, sacrifice their goals
in order to help struggling agents to meet their deadlines. This approach improves optimization considerably
without reducing flexibility substantially, thus guaranteeing higher overall performance and
stability of the system.
16.4.3 Comparison
From a general point of view PABADIS’PROMISE, PROSA, and PABADIS can be summarized as
follows:
• Autonomy and aggregation: On the one hand PROSA lacks flexibility due to the direct control
over aggregated entities and, on the other hand, PABADIS lacks efficiency because of total distribution.
PABADIS’PROMISE defines production order decomposition that provides rules for
controlling autonomous entities without dramatically reducing flexibility.
• Cooperation and hierarchy: Where PROSA has explicit hierarchy that causes rigidness of order
changes, PABADIS provides no hierarchy that implies overhead in the management of complex
products. PABADIS’PROMISE offers an implicit hierarchy (production order decomposition;
flexible structure of orders; dynamic control of resources by resources) that finds the balance
between the first approaches.
• Decision making: While decision-making in PROSA is centralized, meaning there is one control
entity per functionality, PABADIS supports completely distributed decision-making mechanism.
PABADIS’PROMISE provides a semi-distributed approach based on clustering of resources at the
shop floor and production order decomposition at the MES layer.
© 2011 by Taylor and Francis Group, LLC