EnGENEer:

Logica Cambridge Ltd. developed EnGENEer as an in-house GENETIC ALGORITHM environment to assist the development of GA applications on a wide range of domains. The software was written in C and runs under Unix as part of a consultancy and systems package. It supports both interactive (X-Windows) and batch (command- line) modes of operation.

EnGENEer provides a number of flexible mechanisms which allow the developer to rapidly bring the power of GAs to bear on new problem domains. Starting with the Genetic Description Language, the developer can describe, at high level, the structure of the ``genetic material'' used. The language supports discrete GENEs with user defined cardinality and includes features such as multiple CHROMOSOMEs models, multiple SPECIES models and non-evolvable parsing symbols which can be used for decoding complex genetic material.

The user also has available a descriptive high level language, the Evolutionary Model Language. It allows the description of the GA type used in terms of configurable options including: POPULATION size, population structure and source, SELECTION method, CROSSOVER and MUTATION type and probability, INVERSION, dispersal method, and number of OFFSPRING per GENERATION.

Both the Genetic Description Language and the Evolutionary Model Language are fully supported within the interactive interface (including online help system) and can be defined either "on the fly" or loaded from audit files which are automatically created during a GA run.

Monitoring of GA progress is provided via both graphical tools and automatic storage of results (at user defined intervals). This allows the user to restart EnGENEer from any point in a run, by loading both the population at that time and the evolutionary model that was being used.

Connecting EnGENEer to different problem domains is achieved by specifying the name of the program used to evaluate the problem specific FITNESS function and constructing a simple parsing routine to interpret the genetic material. A library of standard interpretation routines are also provided for commonly used representation schemes such as gray-coding, permutations, etc. The fitness evaluation can then be run as either a slave process to the GA or via a standard handshaking routines. Better still, it can be run on either the machine hosting the EnGENEer or on any sequential or parallel hardware capable of connecting to a Unix machine.

For more information, contact: George Robbins, Systems Intelligence Division, Logica Cambridge Ltd., Betjeman House, 104 Hills Road, Cambridge CB2 1LQ, UK. Tel: +44 1716 379111, Fax: +44 1223 322315 (Unverified 8/94).