Information Technology Reference
In-Depth Information
4.1
Manufacturing Problem Setting
For each of the 58 trials run, we obtained a new set of randomly chosen parameter val-
ues, as specified in Table 2. Each paired trial used the same set of parameter values, and
compared standard auctions and AON auctions with the quoting alternatives discussed.
Ta b l e 2 .
Settings of the manufacturing scenario used for our experiments. Parameters specifying
a range are drawn from a uniform distribution. (*) parameter specifies total for all parts in a
configuration, each part getting a random proportion.
Parameter
Va l u e s
General
# of agents (
N
)
4
#ofparts(
M
)
4
Public
I
r
[1000, 2000]
information
D
r
[2000, 6000]
L
[250, 300]
Private
FC
i
[300000, 400000]
information
LC
i
[15000, 20000]
for agent
VC
i
[250,350]
i
|
CF
i
|
2
For
PC
f
[20,60] (*)
each
RC
f
[400000, 800000]
f
∈
CF
i
RT
f
[5,15]
4.2
Agent Bidding
Agents bid in a set of auctions
G
, each corresponding to a different good
r
. Each agent
follows an incremental bidding approach similar to the one described by Cheng and
Wellman [2]. The main loop that controls an agent's behavior is as follows.
1:
repeat
2:
Get price quotes.
Get transactions (i.e., matching bids).
3:
G
do
5:
Select a new point to be added to the bid in
g
.
6:
Fix inconsistencies in bid.
4
7:
Submit updated bid to
g
.
8:
end for
9:
until
Timeout
for
each auction
g
∈
4:
}
The results described in Section 4.4 were obtained by using the same agent structure,
with some variations in terms of selection of new bidding points which are explained
below.
{
allocation process is over
4
Make smallest possible changes to the old points in the bid in order to maintain divisible prices
nonincreasing in quantity and indivisible payments nondecreasing in quantity.
