Neural networks in space

GREENBELT, Md. - NASA scientists are exploring a new concept of distributed spacecraft operations that could revolutionize space science in much the same way distributed computing revolutionized data processing. The enabling factor in each case is the inevitable trend toward ever-smaller, more functionally dense electronics.

Oct 1st, 1998

by John Rhea

GREENBELT, Md. - NASA scientists are exploring a new concept of distributed spacecraft operations that could revolutionize space science in much the same way distributed computing revolutionized data processing. The enabling factor in each case is the inevitable trend toward ever-smaller, more functionally dense electronics.

Dennis Andrucyk, head of an ad hoc earth science vision team at the space agency`s Goddard Space Flight Center in Greenbelt, Md., likens this concept to a web of sensors integrated as an advanced neural network. "It is analogous to the human nervous system in sensing and responding to stimuli, serving as an extension of the human senses and intelligence," Andrucyk told a conference on military and aerospace applications of programmable devices last month at Goddard.

His underlying message is a call for small inexpensive spacecraft linked together in webs. In practice, this concept would challenge the all-the-eggs-in-one-basket approach of the space shuttle since it began operations in 1981. This is not surprising since that is also the mantra of NASA Administrator Daniel Goldin, who has been busy slashing NASA`s budget each year and yet showing more results.

The idea of small spacecraft is not new. Thirty years ago NASA experts successfully launched a series of small scientific payloads under their University Explorer program using low-cost expendable launch vehicles to orbit single experiments.

This idea was particularly appealing to university faculty members The short lead time to launch enabled them to involve their graduate students in the experiments before they received their graduate degrees and wandered away. Participants included the University of Iowa, Rice University, and the University of Rome. What killed the idea was the need to keep the shuttle full, resulting in escalating costs and ever-longer lead times.

On the eve of the Military and Aerospace Applications of Programmable Devices and Technologies conference, NASA resurrected the idea, now known as the University-class Explorers (UNEX) program, and selected the University of California at Berkeley and the University of Minnesota to be the first to launch their experiments.

The payloads are to go into orbit in 2001. The ground rules for the UNEX program, which Goddard is managing, limit the total cost to NASA to $13 million per mission. The idea is to provide frequent flight opportunities for highly focused science missions.

Andrucyk`s ideas go far beyond this one-spacecraft, one-experiment approach. He envisions constellations of spacecraft cooperating in the gathering, processing, and disseminating of data from space, including observations of terrestrial phenomena. He even envisions constellations of satellites operating at the Lagrange points - the regions in space where the gravitational forces of Earth, sun, and moon partially cancel each other out to create particularly stable orbits.

The parallels with the Internet and the evolution of computers from standalone machines to nodes in a global information environment are obvious. We no longer think of one-telephone, one-user or one-computer, one- application. The acceptance of networking has been so widespread as to be beyond our notice.

The first of what Andrucyk calls "micro-observatories," weighing about 220 pounds and costing less than $1 million each, could be launched in five years. Within 25 years he looks for 11-pound spacecraft costing less than $100,000 each to be orbited in constellations of hundreds (and eventually thousands) of spacecraft.

The basic system architecture is to stress autonomous operations using advanced electronics to enable the spacecraft to adapt and reconfigure itself to changing conditions. Andrucyk uses a biological metaphor: they would almost be like living organisms, "growing tentacles [to expand operations] and then biodegrading" over time. This is essentially how adaptive electronics systems operate now to achieve greater reliability through "fail soft" and "graceful degradation" modes.

The short list of necessary supporting electronics technologies includes programmable logic devices to support constantly changing configurations and ever-greater communications bandwidths. This also means lower-power, radiation-resistant devices and advanced sensors, such as focal plane arrays. The present X- and Ka-band phased arrays for "databases in the sky" will eventually evolve into optical communications, says Andrucyk, who expects holographic memories capable of containing petabytes (quadrillions of bytes) of data to be available for space operations in 10 to 15 years.

On Sept. 14 Virginia Gov. Jim Gilmore unveiled a new commercial space launch pad at the NASA field center at Wallops Island, Va. He did this on the day before the programmable device conference - and three days after NASA reinstituted the University Explorer program.

Wallops Island is where the space agency launched the original Explorer satellites using the now-defunct low-cost Scout launch vehicle. To be known as the Virginia Space Flight Center, or VSFC, the installation is scheduled to be ready for operations by mid-2000 with a 200-foot-high service tower and a cleanroom for payload checkout and integration. Experts at the facility will support expendable launch vehicles such as the Orbital Science Corp.`s Taurus series, Lockheed Martin`s Athena I and II, Boeing`s Delta IV, and the U.S. Air Force`s Orbital/Suborbital program Minuteman. NASA officials say they hope to launch seven to nine satellites between 2000 and 2002.

These launch support operations are germane to the electronics industry because industry leaders must anticipate these needs with ever more powerful components if NASA applies the concept of many small, lower cost satellites.

Andrucyk describes programmable logic devices as "the key player ... essential to our vision," and this represents an opportunity for the electronics industry. Money that launch costs ate up in the past should become increasingly available for the payloads. Although NASA officials have never been able to fund improvements to basic electronics technologies, relying instead on the Defense Department during the days of big military spending, the point is moot now that commercial applications are driving the technologies.

The technology base should be the least of the barriers to a new era of space exploration. The bigger challenge may be to change the thinking of the neural networks on Earth who will have to design the proposed neural networks in space.

More in Home