F. Bouille : The Use of G.I.S. in Petroleum Industry
249
algorithmic or heuristic, as complex as required. Moreover, the neural engme
allows automated learning.
In the area of GIS, neural models are particularly convenient for very large models
dealing for instance with pollution propagation, resource distribution, information
transmission, and more generally wherever an automated learning is needed.
Everything modeled with processes can be done with neurons; but both tools are
not strictly equivalent. Processes are more convenient when times have to be
managed, whereas neurons are better suited to the automated searching of one
solution (not simply solution ... ) (see Bouille, 1993b).
The new (not so new ... ) fashion is constraint programming. This consists of
expressing the constraints between several types; a very simple example will
present the problem. For instance, in a first thermodynamic simplified approach, P,
Y, and T are linked by a simple relation which is P*YIT = constant. Try to explain
that to your database! Give P, no problem; give Y, no problem; but now T is
fixed ... You could have first given T, etc .... If you define rules with an expert
system, you need several rules and it will be neither elegant nor efficient. To write
a procedure and call it every time will go back to prehistoric times. You need just
express the formula as a constraint. Of course, constraints in an 0.-0. context, are
objects of classes of constraints managed by a special engine ...
In the GIS area, many constraints can be expressed; to use again the example of
the toponymy automated positioning, the use of constraints could largely improve
the method presently based on rules (this being still very fast). The constraints are
particularly interesting when dealing with superimpositions, intersections and
interconnections of networks, mainly in urban areas.
2.3 The devices for GIS: a revolution in use
The era of multimedia has changed most uses, and mainly minds; a simple
example in this brief paragraph: considering a complex urban network, you have
to deal with urban furniture; when pointing to an object of a class of furniture, as a
telephone cabin, you may have the image of this cabin, or you may have some
associated plans. In an environmental application, concerning the distribution of
birds, when pointing to an object of a class of birds, you may hear its song, and
simultaneously you may see its image, or for expressing the way it flies, you may
see a brief movie.
Nowadays, cartographic robots, or exploring robots in hostile environments, are
very useful and very numerous; thousands of robots work for us, generally on
materialized networks, for instance, in oil exploration and production industry, for
cleaning pipes; or for controlling the position of pipes under the sea; or for
digging under the sea, welding pipes, burying pipelines, etc .... A GIS must be able
to be connected to such applications.
Among other real-time operations, requiring high speed, a GIS can be used for
cars wandering through all streets of a town and emitting radio signals which
launch local transmitters, providing immediate and automated gas-meter readings
without stopping cars and without entering houses. Many new applications like this
will be launched in the next decade, including the remote controlled vehicles.
249
algorithmic or heuristic, as complex as required. Moreover, the neural engme
allows automated learning.
In the area of GIS, neural models are particularly convenient for very large models
dealing for instance with pollution propagation, resource distribution, information
transmission, and more generally wherever an automated learning is needed.
Everything modeled with processes can be done with neurons; but both tools are
not strictly equivalent. Processes are more convenient when times have to be
managed, whereas neurons are better suited to the automated searching of one
solution (not simply solution ... ) (see Bouille, 1993b).
The new (not so new ... ) fashion is constraint programming. This consists of
expressing the constraints between several types; a very simple example will
present the problem. For instance, in a first thermodynamic simplified approach, P,
Y, and T are linked by a simple relation which is P*YIT = constant. Try to explain
that to your database! Give P, no problem; give Y, no problem; but now T is
fixed ... You could have first given T, etc .... If you define rules with an expert
system, you need several rules and it will be neither elegant nor efficient. To write
a procedure and call it every time will go back to prehistoric times. You need just
express the formula as a constraint. Of course, constraints in an 0.-0. context, are
objects of classes of constraints managed by a special engine ...
In the GIS area, many constraints can be expressed; to use again the example of
the toponymy automated positioning, the use of constraints could largely improve
the method presently based on rules (this being still very fast). The constraints are
particularly interesting when dealing with superimpositions, intersections and
interconnections of networks, mainly in urban areas.
2.3 The devices for GIS: a revolution in use
The era of multimedia has changed most uses, and mainly minds; a simple
example in this brief paragraph: considering a complex urban network, you have
to deal with urban furniture; when pointing to an object of a class of furniture, as a
telephone cabin, you may have the image of this cabin, or you may have some
associated plans. In an environmental application, concerning the distribution of
birds, when pointing to an object of a class of birds, you may hear its song, and
simultaneously you may see its image, or for expressing the way it flies, you may
see a brief movie.
Nowadays, cartographic robots, or exploring robots in hostile environments, are
very useful and very numerous; thousands of robots work for us, generally on
materialized networks, for instance, in oil exploration and production industry, for
cleaning pipes; or for controlling the position of pipes under the sea; or for
digging under the sea, welding pipes, burying pipelines, etc .... A GIS must be able
to be connected to such applications.
Among other real-time operations, requiring high speed, a GIS can be used for
cars wandering through all streets of a town and emitting radio signals which
launch local transmitters, providing immediate and automated gas-meter readings
without stopping cars and without entering houses. Many new applications like this
will be launched in the next decade, including the remote controlled vehicles.