Genes to form basis of space missions - NASA's Darwinian design modelling technique - Government Activity

Electronics Times, April 26, 1999 by Matthew Alleyne

Nasa scientists have developed a novel approach to the design of space missions based on forming a `gene pool' from data compiled in previous space flights from which future missions will evolve.

The modelling technique, referred to as Darwinian design, uses a computer that applies Darwinian principles to track the evolution of previous missions. The compiled data is then used to create missions based on the best traits of previous space flights.

The concept was presented in a paper by Dr David Noever and Dr Subbiah Baskaran, both of Nasa's Marshall Space Sciences Laboratory.

According to Dr Noever: "The next generation of spacecraft will more than likely evolve traits that their parent ships could not have equalled."

Nasa says, although computers can create infinite lists of combinations to try and solve particular problems, users don't want a computer to "spew out" endless random lists of possibilities.

Instead, by breeding the most successful operations, successive generations can learn from past mistakes, and improve.

This idea of survival of the fittest gives rise to the Darwinian reference, and is one of the key ingredients in artificial intelligence when applied to computer design.

Noever said: "What it boils down to is trying to get computers to recognise patterns and react. We treat the design choices of a mission like a biologist treats a genetic chromosome. We can cut and splice missions together with remarkable speed, compete them against each other, and then multiply the survivors."

Nasa has used the approach on recent missions such as the Mars Pathfinder, the Earth Observer and the Deep Space 1 interplanetary probe.

Noever and Baskaran are particularly excited about the Deep Space 1 mission as it is the first major spacecraft expected to "learn" during its trip through the solar system.