Information Technology Reference
In-Depth Information
542 Advanced Artificial Intelligence
to look at how tokens would be passed if it were feasible. Then, by dividing the
monolithic joint activity into a set of actions that can be taken by individual
agents, we can decentralize the token routing process where distributed agents, in
parallel, make independent decisions of where to pass the tokens they currently
hold. Thus, we effectively break a large coordination problem into many small
ones.
Algorithm 14.7
Token pass local model which makes decision process for
agent α to pass incoming tokens (Xu et al,2005).
Method:
(1) while true do
(2)
Tokens
(α) ←
getT oken
(
sender
);
(3) for all ∆ ∈
Tokens
(∈) do
(4) if
Acceptable
(∆,
α) then
(5) if ∆.
type
==
Res
then
(6)
Increase
(∆,
threshold
);
(7) end if
(8) else
(9)
Append
(
self,
∆
.path
);
(10) for all ∆
i
∈
H
α
do
(11)
Update
(
P
α
[∆],∆
i
);
(12) end for
(13) if (∆.
type
==
Res
)
‖
(∆.
type
==
Role
) then
(14)
Decrease
(∆.
threshold
);
(15) end if
(16)
acquaintance
←
Choose
(
P
α
[∆])
(17)
Send
(
acquaintance,
∆);
(18)
AddtoHistory
(∆);
(19) end if
(20) end for
(21) end while
In above model Pα is the decision matrix agent α uses to decide where to
move tokens. Initially, agents do not know where to send tokens, but as tokens
are received, a model can be developed and better routing decisions made. That
is, the model, Pα is based on the accumulated information provided by the
receipt of previous tokens. Algorithm 14.8 shows the reasoning of agent α when
it receives incoming tokens from its acquaintances via function getToken(sender)
(line 2). For each incoming token ∆, function Acceptable(α, ∆) determines
