This technique uses simulated virtual-world evolution, incorporating the principles of natural selection, to solve a range of non-linear mathematical problems. However, in GAVONTS, it is a population of non-linear mathematical vector transformations, rather than organic life, that are adapted, selected and evolve.

This technique is a 'meta-model', because it is an algorithm, or mathematical model, which itself generates mathematical models. The models that are generated being selected according to how well they solve given 'target' problems.

The 'genetic' element of this approach is related to the way in which these transformations are structured, in such a way as to facilitate sensible functional combination strategies. This enables us to 'breed' these functions, creating new ones which can 'inherit' the characteristics of their parents.

See the: GAVONTS page in our consultancy services area.

GAVONTS Implementation - Using XML-Space

In order to process the vast number of calculations that are often required to run GAVONTS, we use our xml-space parallel processing technology. The GAVONTS framework can be run in a number of possible configurations, according to the problem being tackled.

For example, in the case of the 'Lander Example', the architecture that was used consisted of a number of quasi-isolated 'island' populations of GAVONTS solutions, each evolving semi-independently. These populations are weakly coupled to one-another, through the infrequent random exchange of individuals, via an intermediating xml-space.

As these different populations evolve, xml-space message channels are also used, to enable selected individuals within these populations to be exported from their islands for logging, persistence, and further analysis.

Applications and Examples

This technique has a wide range of potential applications in a number of different areas. We have included an example on this site - where GAVONTS has been used to create a control system algorithm that is able to land a spacecraft on a planetary surface successfully, in simulation.

Performing such a landing is not a straightforward thing to do, and the control algorithm that was generated by GAVONTS, without any human guidance, is able to do this consistently from a range of starting configurations, and with almost pinpoint accuracy.

See the: GAVONTS Lander Example page.