INTELLIGENT CONTROL OF MATERIAL HANDLING SYSTEMS - Environmentally Conscious Materials Handling - page 100

C ' s domain

C ' s domain C ' s domain

U 18 U 17 U 16 U 15 U 14 U 12 U 11 U 10 U 9 U 8 U 7 U 6 U 5 U 4 U 3 U 3 U 1

U 13

C 3 C 2 C 1

Input buffer C 0

Figure 3.24 Generalized overview of the system for the anodization process, with bars moving from right

to left

frame for processing each step of the recipe complicates the task. 5 Another issue
is the dynamical production environment, which has great impact on the system's
ability to recover and finish the current bars while running as best as possible
under partial breakdowns. Examples of unpredictable error conditions could be
that the temperature of a bath is too low and must be heated before being avail-
able again, cranes break down, the liquid level of a bath is too low, or orders
are too rapid. An agent-based approach must focus on the dynamics and be able
to recover or continue as best as possible under such conditions.
The problem is classic — the throughput of a system should optimize the flow
between subprocesses and handle the inflow process correctly to best utilize the
system.
In abstract terms, there exists a number of tasks, q i 1 , 2 ,...,n ,

with k i subtasks 6 q i, 1 ,q i, 2 ,..., q i,ki . The subtasks are interconnected, and the
order cannot be changed. Tasks should be handled as visits to processing
stations — determined by the recipe.

for i =

12.1 PACO Approach

PACO is a contraction of coordinated patterns (Demazeau 1991) and takes a sim-
ple approach of designing the agents. PACO focuses on reactive agents situated

5 Only minimum and maximum times are given for each step of the recipe.
6 Note that the number of subtasks might be different for each taskgroup.