Evolution optimizes satellite orbits

By Kimberly Patch, Technology Research News

Plotting orbits for a constellation of satellites is a complicated business.

Orbiting closer to earth makes it possible for a satellite to communicate with low-power devices like cellphones, but doing so produces more blind spots, because to maintain contact a satellite must have a line of sight to earthbound antennas.

Researchers from Purdue have used genetic algorithms to design orbits for constellations of satellites that are more efficient than current group orbits. Conventional orbits for constellations of three or four low-altitude satellites that circle the earth in about 90 minutes can cover most of the earth for most of the day, but have blackout periods.

Satellite orbits are tricky to plan because they have two objectives: to shorten both the longest gap in coverage and the average time a gap will last, said William Crossley, an associate professor of aeronautics and astronautics at Purdue University.

The trouble is, shortening one type of gap often widens the other. "The optimization routine [that] concentrates on reducing the coverage gaps that are the longest in duration... does not consider the average revisit time experienced by all ground stations. Conversely, designing a constellation to minimize [average coverage gaps] will reduce the revisit time for most points in the area of interest on the earth, but this may allow some locations to have very long gaps in coverage," said Crossley.

The researchers found a more efficient orbit using a type of genetic algorithm that can handle two objectives. Genetic algorithms are based on the Darwinian model of natural selection, or survival of the fittest. "In the computer, a population of possible satellite constellations is represented by a chromosome," said Crossley.

The researchers' satellite chromosome included three types of orbit information: the orbit's angle to the equator, the longitude where the satellite orbit crosses the equator, and how far from the equator each satellite is when the first satellite in the constellation crosses the equator. Each constellation also had a fitness value depending on the efficiency of its combination of variables.

The genetic algorithm mixed the chromosomes of the most fit individual constellations to make a new generation of constellations, and the fittest ones went on to produce new generations. "The survival of the fittest behavior acts to improve designs over numerous generations," Crossley said.

The researchers ran the satellite algorithm for 200 generations in order to generate large numbers of solutions mapping the trade-offs in minimizing the two types of gaps. "We conducted these investigations for different numbers of satellites in different altitudes, and the genetic algorithm generated between 10 and... 30... designs representing the trade-offs," said Crossley.

An interesting pattern emerged. "We found that for small numbers of satellites, the constellations with low [gaps] were often nonsymmetric," meaning that there are nonuniform intervals between the satellites' equator longitude and equator distance values, Crossley said. "The nonsymmetric constellation results were fairly surprising, mostly because traditional [orbit mapping methods] relied heavily upon symmetric orbits," he said.

Crossley is applying the approach to other areas, while his research colleagues at the Aerospace Corporation are using the satellite data to find practical constellations, Crossley said. In general, the approach can "help engineers and designers search through a complex design space and find good, possibly nonintuitive solutions to aerospace problems," he said.

The research does a good job of applying multi-objective genetic algorithm methods to an interesting problem, said Erik Goodman, a professor of electrical and computer engineering, and mechanical engineering at Michigan State University, and vice president of technology at Applied Computational Design Associates, Inc. "This sort of planning, and particularly the explicit presentation of the trade-offs between the two objectives... is difficult to do" using conventional methods, and is therefore a good application for this type of genetic algorithm, he said.

Crossley's research colleagues were Edwin A. Williams of Purdue University and Thomas J. Lang of the Aerospace Corporation, a nonprofit entity that provides science and engineering consulting services to the U.S. Air Force and other U.S. government entities.

The research was funded by the Aerospace Corporation.

Timeline:   Now
Funding:   Private
TRN Categories:  Artificial Life and Evolutionary Computing; Applied Computing
Story Type:   News
Related Elements:   Related technical paper, "Average and Maximum Revisit Time Trade Studies for Satellite Constellations Using a Multiobjective Genetic Algorithm," AAS/AIAA Space Flight Mechanics Meeting in Clearwater, Florida January 23-26, 2000.


November 14, 2001

Page One

Crossed nanowires compute

Disappearing links shape networks

Stored light altered

Flipping flakes change color

Evolution optimizes satellite orbits


Research News Roundup
Research Watch blog

View from the High Ground Q&A
How It Works

RSS Feeds:
News  | Blog  | Books 

Ad links:
Buy an ad link


Ad links: Clear History

Buy an ad link

Home     Archive     Resources    Feeds     Offline Publications     Glossary
TRN Finder     Research Dir.    Events Dir.      Researchers     Bookshelf
   Contribute      Under Development     T-shirts etc.     Classifieds
Forum    Comments    Feedback     About TRN

© Copyright Technology Research News, LLC 2000-2006. All rights reserved.