Information Technology Reference
In-Depth Information
and the reference price modified by a factor related to the requested delivery date. This
ensures the agent sells the PC at least for its cost. The use of
d
due
means that the sooner
the due date, the higher the offered price is compared to the reference price (because
the agent has little time to produce the computers with a bigger risk of being penalised
for being late).
In more detail, the fuzzy reasoning inference mechanism employed to set the adjust-
ment factor in Equation (1) is based on the standard Sugeno controller [8]. Our agent
uses two rule bases: one for the end stage of the game (about last 40 days) and one
for other days in the game. Both rule bases incorporate some 20 rules
4
which vary the
price according to the market demand, its inventory level and time into the game (see
Appendix for details of the rule bases). We show two representative rules below:
R
j
:
if
D
is
high
and
I
is
high
and
ND
is
far
then
r
j
is
big
R
q
:
if
D
is
high
and
E
is
close
then
r
q
is
very
-
small
where the customer demand (
D
) is expressed in the fuzzy linguistic terms
high
,
medium
,
and
low
, the inventory level (
I
)intheterms
high
,
medium
,and
low
, next delivery date
of a big amount of components (
ND
)intheterms
far
,
medium
,and
close
and days to
the end of the game (
E
)intheterms:
far
,
medium
,and
close
.
r
j
is the output of the
individual rule
j
(
i.e.
, the adjustment factor discussed above). Thus, rule
R
j
captures
the fact that if the type of PC is in high demand in the market, the agent has a high
inventory for this kind of PC and there is a long time until the next delivery for a high
volume of components, then
r
j
should be big (thus resulting in a higher bid price). The
second rule
R
q
captures the fact that if there is high demand for a particular type of PC
and there is a little time until the end of the game, then
r
q
should be very small (thus
ensuring a low offer price and hence reducing the risk of being left with inventory at
the end of the game). The firing level
α
j
∈
R
j
is computed in the standard
way by using the
Min
operator on the membership values of the corresponding fuzzy
sets. According to the Sugeno controller definition, the crisp control action (
i.e.
,the
output of the fuzzy rule base fed into Equation (1)) is:
[0
,
1] of rule
j
=1
α
j
r
j
j
=1
α
j
r
=
(2)
Adaptation of offer prices.
Given the uncertainty in TAC SCM, we believe it is es-
sential for the agents to be responsive to the prevailing situation during the course of
bidding for customer orders. The idea is that the agent can only use 2000 production cy-
cles every day, so, to maximise throughput, the number of cycles necessary to produce
the received customer orders should also be 2000. Thus if the received orders require
4
In generating the rules, we followed the steps below: (1) determine what to reason about
in this SCM game - the offer price; (2) choose the factors that should be used in the rules -
inventory level, demand, and time into the game (there may be some other factors, but these are
the most relevant ones); (3) structure the fuzzy rules - based on the relationship of the factors
to the reasoning value and experiences in the field; (4) decide how to adapt the parameters in
the rules - SouthamptonSCM adapts the price based on the quantity of the received customer
orders and the expected number of orders. (5) refine the rules and parameters.
