|
The key difference
of B2B-OS compared to many other agent-based
marketplace projects is the absence of a central arbitrator or auctioneer,
and the implementation of a strictly decentralized
automated negotiation (bargaining) protocol. The goods that are
traded are defined as commodities, so the only negotiation variable is
the price.
The competition
between agents of the same type leads to an individual
agent strategy, where concession making and greediness is in some
sort of balance as a functional result of nothing but the strategies of
the other agents in the population.
The B2B-OS agents use an adaptive strategy based
on a non-deterministic finite state automaton (non-det FSM), in
which action paths are taken depending on stochastic probes against certain
internal parameters. It is not a goal of B2B-OS to simulate real human
behavior, but to compare future automated negotiation strategies, which
are expected to be mostly derived from deterministic rule-based models
with only a handful of variables. |