Information Technology Reference
In-Depth Information
An Artificial Market for Efficient Allocation of Road
Transport Networks
Matteo Vasirani and Sascha Ossowski
Centre for Intelligent Information Technology
University Rey Juan Carlos, Madrid, Spain
{
matteo.vasirani,sascha.ossowski
}
@urjc.es
Abstract.
The efficient utilisation of large and distributed socio-technical
systems is a difficult problem. Centralised approaches can be computationally
intractable, unresponsive to change and require extensive knowledge of the un-
derlying system. In the last decade, distributed approaches based on artificial mar-
kets have been proposed as a paradigm for the design and the control of complex
systems, such as group of robots or distributed computation environments. In this
work we model an artificial market as a framework for the efficient allocation of
a road transport network, where each network portion is controlled by a market
agent that “produces mobility” on its links. We demonstrate that the collective
behaviour of the market agents, if properly designed, lead to an optimised use of
the road network.
1
Introduction
Achieving an efficient utilisation of large and distributed socio-technical systems is a
difficult problem. In general, centralised approaches tends to be computationally in-
tractable, unresponsive to change and require extensive knowledge of the underlying
system. Distributed approaches are not as prone to these problems, but they can be
highly sub-optimal. In the last decade, economic approaches based on artificial mar-
kets [1] have been proposed as a paradigm for the design and the control of complex
socio-technical systems, such as group of robots [2] or distributed computation envi-
ronments [3][5].
In this work we propose the application of an artificial market for the efficient alloca-
tion of a road transport network, where each network portion is controlled by a market
agent that “produces mobility” on its links. We model the price selection strategy of
the market agents in such a way that the overall outcome of the market is aligned with
the minimisation of the social transportation cost. We analytically demonstrate that the
collective behaviour of the market agents leads to an optimised use of the road net-
work. Finally, we define a learning-based framework for the application of the artificial
market in real-world road transport scenarios, when demand and cost are dynamic and
uncertain functions.
This paper is structured as follows: in section 2 we introduce our notion of artificial
market and its application to a simple road transport scenario. In section 3 we go beyond
the pure economic analysis and we extend the artificial market to be applied to real-
world road transport networks. In section 4 we perform an experimental evaluation.
Finally we conclude in section 5.
