Source: AI Expert, April 1993
Prolog is no longer confined to research laboratories, but is now considered to be a powerful tool for the development of commercial applications
Prolog is no longer the best-kept secret in the programming industry. The Prolog programming language is not solely an AI language confined to research laboratories, but is increasingly seen as a powerful tool for the development of practical applications. The technology is now in widespread use in a wide range of domains, including finance, defense, telecommunications, law, medicine, agriculture, engineering, manufacturing, and education. Prolog has some clear advantages over other programming languages and can to significant returns on investments and completive advantages. Through the recently formed Prolog Vendors Group together with events such as the Prolog 1000 catalog, we now have solid information that many hundreds of fielded applications, demonstrating that Prolog is fast becoming the language of choice for many software developers.
HISTORY
The Prolog programming language originated from work done in the early 1970s
by Robert A. Kowalski at Edinburgh University (and ever since at Imperial College,
London) and Alain Colmerauer at the University of Aix-Marseille (France). Kowalski's
research on the field of automated deduction led in 1972 to the development
with Colmerauer of the use of formal logic as a programming language. Kowalski's
research provided the theoretical framework, while Colmerauer's gave rise to
the programming language Prolog. Colmerauer and Phillipe Roussel built the first
Prolog intrepreter, and David Warren at the University of Edinburgh built the
first Prolog compiler (for the DEC-10 machine). Most commercial implementations
of Prolog now use this same Edinburgh Prolog syntax.
Since then, Prolog has grown in popularity throughout Europe, the United States,
and in Japan. In Europe, Prolog is widely used. In the United States, Prolog
is thriving, and suppliers of Prolog-based constraint technology report that
it is a prime market. In Japan, Prolog was the fundamental technology underlying
the ICOT Fifth Generation Computer System Imitative, which started in 1982.
The Japanese interest in Prolog has gone some way in accounting for the growth
of the language elsewhere; although interestingly, ICOT has been widely regarded
as a major success.
PIGS
When you ask what Prolog is good for, it is tempting to point to its clear success
for a particular set of application areas, including natural language and databases.
While it is true that Prolog has qualities that are especially well-suited to
these domains, Prolog is used for a surprisingly broad range of application
areas.
One of the more interesting applications is PigE/Auspig—and expert system for
raising pigs, which won the award for the best presentation at The First International
Conference on the Practical Applications of Prolog. PigE is an intelligent backend
to a mathematical modeling package called Auspig. The Auspig model simulates
the growth and reproduction of pigs, identifies factors that limit optimal performance
of the pig, and suggest management strategies that maximize profits. PigE presents
an abstracted description of the output of the model in a form a nonmathematician
can understand and suggests dietary, housing, genotyping, or resource input
changes, which the developers say can dramatically improve the profitability
of the herd.
Despite the rather wacky nature of this application, the PigE/Auspig have led
to some startling returns on investment. PigE discovered that increasing the
air flow over a hot pig shed by 0.2-0.6 meter per second (an increase not really
discernable by humans), doubled the animals' growth rate. In one spectacularly
successful study of a pig farm, Auspig was used to develop a growing regime
that led to 496% increase in profits (from $40,992 to $202,231).
According to the developers of the system, verification studies have demonstrated
that the expert system can significantly outperform human experts interpreting
the output of the model (performance measured in dollars per square meter per
day). In an unusual case, the improvement is in the order of 10%. If spread
over the 300 kiloton $500 million (per annum) Australian pig herds, the developers
believe that this improvement would be in the order of $50 million. PigE is
now in routine use in the United States, the Netherlands, Belgium, France, Spain,
and Australia.
ENVIROMENT
Another area where Prolog is successful is as environmental systems that predict
the weather, analyze water supplies, and so on. The MM4 Weather Modeling System,
developed at Penn State University and the National Center for Atmospheric Research
in (Boulder Colo.), gives relatively detailed, short term forecast of meteorological
conditions if the continental United States. It is used particularly to model
pollutant dispersion (for instance, acid rain deposition). The original program,
which contained 16,000 lines of FORTRAN code, was harnessed with Strand88, and
10-fold speedups have been obtained with 15 parallel processors.
The Air Pollution Control System, developed using the Hungarian product MPROLOG,
handles data about the concentration of industrial pollutants in Budapest and
other countries of Hungary. Designed for such users as managers and research
workers, the system checks whether the air pollution of working or planned plants
is below the permitted level. If permitted levels are exceeded, the system recommends
appropriate filtering equipment.
RoadWeather is a 24-hour weather prediction system for snow and ice control
on highways. Developed using PDC Prolog version 3.30 for Windows 3.1, RoadWeather
Pro consists of three components. The first component is a numerical weather
prediction system that forecasts from specified initial (radio sound and surface
observations) and boundary conditions (detailed terrain effects) using a system
of hydrodynamic and thermodynamic differential equations. This part of the system
is written in FORTAN. The second component is a graphical user interface (GUI)
written in Prolog under Windows 3.1, which the developers say is fully interactive
and displays predicted weather phenomena as symbolic objects. The final component
is an Expert Weather Advisor, written in Prolog and Windows 3.1, which permits
mouse point and click manipulation of weather "objects" thereby allowing
forecast upgrades based of recent observational data received from sensors or
human observers. The RoadWeather Pro system is in use at the WELS Research Corp.
(Boulder Colo.).
WATER
A series of Prolog expert systems has been developed at the U.K.'s University
of Surrey, and several are targeted for the water industry. One of these systems
WIESSE, which was built as part of the WIESC Alvey Project, provides an expert
system development environment. WIESSE has been used to develop two expert systems:
WADNES and SERPES. WADNES provides support for emergences that can occur in
a water distribution system by using a model of the distribution network to
suggest methods of re-supplying water—informing appropriate bodies and managing
remedial resources. WADNES creates input data files for large commercial FORTRAN
simulation program, runs the simulation, and interprets the output through intelligent
processing. While these applications were developed as demonstrator systems,
they are generating considerable interest throughout the U.K.'s private water
companies.
Paul Holmes-Higgins, senior researcher fellow and prime mover behind the development
of these systems, believes that Prolog is an ideal language for rapid prototyping,
since it contains characteristics and constructs that would otherwise have to
be implemented in C. "Characteristics such as unification and arbitrary
data structures (Prolog terms) allow me to 'sketch' a program out and yet be
able to execute it; there are no type declarations required, no memory allocation,
just solving the problem," he said. Holmes-Higgins says this lets him involve
the end user at a much earlier stage in the development process.
Holmes-Higgins notes that Prolog is a higher level language than C, so there
is much less source code to maintain. "A well written Prolog program is
more readable than a well written C++ program by respective experienced programmers"
he said.
Historically, Prolog has sometimes been criticized for poor performance, but
many argue that this is no longer a problem. Indeed, Holmes-Higgins believes
that current Prolog technology provides run-time performance that is not noticeably
slower than C programs. Often, says Holmes-Higgins, Prolog applications have
a smaller memory "footprint. "I think commercial Prolog systems have
a smaller 'footprint' in terms of memory and disk requirements than Lisp systems.
You can write some very useful programs in DOS-based Prolog using the 640K limit—I'm
not aware of a 640K DOS Lisp system"
AIR
Boeing Corp., the aircraft giant, is a major user of Prolog technology. Connector
Assembly Specifications Expert (CASEy) is an expert system that guides shop
floor personnel in the correct usage of electrical process specifications. CASEy
provides detailed assembly instructions for terminating wire into electrical
interconnects such as connectors, terminal lugs, splices, and so on. It provides
a prioritized list of authorized tools, settings, materials, and procedures
(with graphics), needed to perform the task. In practice, the system reduces
the time for an operator to assemble a set to instructions from 42 to five minutes.
A Generalized Query System, developed by Boeing Computer Services, uses a client/server
architecture and provides a high-level abstraction mechanism for access to diverse
sources of data and knowledge. The system offers an access mechanism for either
a Prolog or C program that provides a data abstraction service layer. External
data access and interpretation are encapsulated in the conceptual views and
meta data of the Generalized Query Server. The "core" Generalized
Query System software provides a query language and network access layers so
that application developers can access any of the encapsulated data sources.
Data from flat files, databases, object-orientated databases, or even expert
systems can be encapsulated, and the developers say that the query language
makes the physical data access mechanisms and network layers transparent.
The system follows work on "mediators" (by Wiederhold at Stanford
University), or intelligent agents. Bill McClay, senior principal scientist
and AI specialist at Boeing, explains that the idea is to have intelligent agents
between databases and applications in a distributed client-server network. These
give a much higher level of abstraction in manipulating the data inside such
crude things as databases. McClay says "Basically the idea is that you
turn that data into information by having a layer of intelligence in each of
the nodes where the data exists."
The developers say the generalized query language is much like Prolog, but has
several additional features such as a second order logic, abstract data types,
and object oriented syntax. The systems were developed using Quintaus Prolog
version 3.1 running on Sun 4, VAX, and RS6000 workstations.
Bill McClay says it would have been much harder to develop these applications
in any other language. "Prolog was a particularly good match for building
the query engine because the language was so close the semantics of the language
we wanted to build in terms of backtracking and unification." Comparing
Prolog with Lisp and other languages, he added, "The work necessary to
write a query engine in prolog was minimal compared to Lisp. Doing it in a conventional
language would have been unthinkable, because it would have meant re-inventing
Prolog."
BALL GAMES
Prolog is being used for fielded heavyweight applications, and benefits can
result from using the language. But Prolog is a far more interesting language
than, say, C++—and it is starting to finds its way into our homes.
The Fantasy Basketball system has been used for two years to produce the weekly
standings for the Bay Area Fantasy Basketball League of Palo Alto, Calif. Fantasy
(or Rotisserie) Basketball leagues are a growing hobby among avid basketball
fans, consisting of fictitious teams who are in the NBA and featuring competitions
between teams based on the game statistics of their NBA players.
John Dowding maintains Rotisserie/Fantasy Basketball leagues, keeps weekly NBA
statistics in a Prolog database, and ranks NBA players and teams. The Prolog
program downloads the weekly NBA stats, converts them into a Prolog clause database,
and computes the weekly standings for teams in the league. The resulting NBA
statistics database is also used to rank NBA players and predict future performance.
Dowding says Prolog has been an excellent vehicle for this task because of its
built-in support for databases, sorting, and parsing. The system was developed
using Quintus Prolog versions 2.5-3.1.
Toernooi Assistant, developed by J.G. de Lint of the Netherlands, aims to automate
everything about tennis tournaments. Developed using PDC Prolog v. 3.3, Toernooi
Assistant provides optimal planning and scheduling of matches, fast replanning
in case of rain, player administration, seeding and drawing of lots for the
arrangement of the players, processing of the match results, generation of press
releases, and financial support.
Nor is Prolog limited to basketball and tennis. Perfect Pitch is a sports advisor
expert system built to advise the safest and most economical sports surface
to use for a given sport. The developers say that the system will also advise
formal specifications to aid in the tender process. Developed using PDC Prolog
v. 3.21, the system is being used in Australia and is aimed particularly at
local government, engineers, and architects.
THE GOLD RUSH
Finally, the U.S. gold rush could have been a lot more effective using Prolog.
Goldfinder is a knowledge-based system that advises an exploration geologist
on where to find gold. Developed by D.D. Hawkes using MacProlog from the London-based
Logic Programming Associates, the Goldfinder system assesses the potential of
a gold prospect and gives advice on the best location to site a drill hole.
The geologist's map of the prospect is the primary source of input data. Knowledge
is stored partly as semantic networks but mainly in a production rule formalism.
Through interaction with the user, the system establishes a sophisticated dynamic
database containing information about the main lithologies, the nature of any
mineralization or alteration, and the structural features of eh prospect. Using
a method of intersecting loci, the system locates a drill site with a precision
governed only by the scale of the map. Race you to the hills folks..... but
don't bring your C++ compilers.
PROLOG IS EVERYWHERE
Prolog's fast incremental development cycle and rapid prototyping capabilities
have encouraged the use of the language as a tool for solving AI problems. However,
other features of Prolog make it a particularly powerful tool for building robust
commercial applications. These features include interfaces to other languages
and database products, stand-alone application generators, and more recently,
support for techniques such as object-orientated and constraint based programming
Object-oriented extensions to Prolog have increased the attractiveness and expressive
power of the language and provide Prolog users with a powerful and flexible
object-orientated development language. Constraints replace Prolog's usual pattern-matching
mechanism with a more general operation called "constraint satisfaction."
Constraints are a powerful way to reduce the size of the search space and thus
increase the efficiency of the scheduler.
Prolog has grown from a simple stand-alone interpreter into a full compiler-based
technology with links to external databases, GUIs, and other languages such
as C, and in so doing has effectively made the transition from academic-theorem
prover to a complete general-purpose programming language. Prolog technology
will continue to mature and build on its foundation. The language will be increasingly
deployed in systems, data processing systems, and specialized markets such as
scheduling, help-desks, configuration and planning.
Enhancing the declarative nature of logic programming with additional features
such as object-oriented extensions and constraint-based techniques results in
very efficient and powerful Prolog systems. These facilities, together with
increased integration to fourth-generation languages, are helping to position
Prolog as an effective information-management tool. By adding value to mainstream
data processing environments, Prolog will provide the key to handling the complex
information repositories of the future.
Al Roth is a freelance writer and group secretary of the Prolog Vendors Group.
