A small robotic observatory system, called RAPTOR, is poised to take movies of fleeting astrophysical events. These movies will help astronomers better understand planetary systems, stars, galaxies, and the universe. Some of RAPTOR's data analysis techniques can also be applied to defense problems.

In January 23, 1999, a NASA satellite detected a brilliant burst of gamma rays 10 billion light years from Earth. The satellite transmitted the burst's position to a small robotic observatory at Los Alamos National Laboratory called ROTSE, for Robotic Optical Transient Search Experiment. Within 10 seconds, ROTSE pointed its telescopes at the burst and began taking a movie of the most luminous celestial object ever observed. This was also the first time the light from a gamma-ray burst had been recorded as the burst emitted gamma rays. The pulse of light lasted about 80 seconds.

Many distant astrophysical events also produce pulses of light, including exploding stars, extrasolar planets, stellar flares, binary star systems, pulsating stars, and gravitational microlensing events. These "transient optical events" provide important information about the universe. For example, gravitational microlensing studies can help determine the composition of "dark matter," which makes up 96 percent of the universe's mass but emits no light and is therefore invisible. Dark matter in the early universe provided the clumps of mass needed to form galaxies, stars, and planets. (Gravitational microlensing occurs when the gravity of a dark massive object passing between Earth and a star focuses the star's light to make it brighter as seen from Earth. Microlensing objects include brown dwarf stars and black holes.)

To capture a transient optical event, however, a telescope must center the event in its field of view before the event disappears. The massive telescopes of conventional observatories move too slowly, but small, computer-controlled telescopes are nimble enough for the task.

Last fall, a new system of robotic observatories became operational at Los Alamos. Called RAPTOR, for Rapid Telescopes for Optical Response, the system took its first movie of a gamma-ray burst in response to a satellite alert on December 11, 2002. [figure: first gamma-ray movie]

However, RAPTOR can do more than respond to satellite alerts. Equipped with sophisticated computer intelligence, it is the first robotic observatory system that can find and study transient optical events on its own. It is also the only robotic observatory system with stereovision, which allows it to discern between transient optical events and nearby space junk, as well as to detect "killer" asteroids (see the sidebar: Detecting "Killer" Asteroids).

A New Window on the Universe

RAPTOR could also be the first observatory to take movies of such exotic objects as giant flares on sunlike (solar) stars and orphan gamma-ray bursts. Although these objects are thought to exist, they have been difficult to observe: the few sightings of giant solar flares are in doubt, and orphan bursts have not yet been seen. Only RAPTOR has the intelligence and speed to identify and capture these fleeting events.

And there are good reasons to study such events. A giant flare on the sun could destroy Earth's climate and its inhabitants. Even an ordinary "small" flare can affect the weather, overload power grids, and knock out satellites. Studies of giant flares on other solar stars could help astronomers predict the likelihood of such a flare on our sun.

Orphan gamma-ray bursts are of interest because at least some gamma-ray bursts are caused by exploding stars, which seed the universe with the heavy elements of which planets—and people—are made. Studies of gamma-ray bursts will elucidate how stars explode (see the sidebar on gamma-ray bursts).

System Overview

RAPTOR was built by a Los Alamos team headed by astrophysicist Tom Vestrand; the project was funded by the Los Alamos Laboratory Directed Research and Development Program. The system consists of four small robotic observatories. RAPTOR-A, -S, and -P are at Fenton Hill, in the Jemez Mountains west of Los Alamos. RAPTOR-B is at the Los Alamos Neutron Science Center, 38 miles away. Computers located between the two sites analyze data from the observatories and send commands to computers at the sites that control the observatories' telescopes and digital cameras. The computers communicate through the Internet.

RAPTOR-A and -B are identical and together provide RAPTOR's stereovision. Each observatory consists of a wide-field telescope and a narrow-field telescope mounted on a platform that can swivel to any point in the sky in less than 3 seconds. The telescope platform is the fastest ever built. [figure: RAPTOR-A telescope]

With a field of view of 38 by 38 degrees, each wide-field telescope can cover the sky in eleven patches. However, RAPTOR usually focuses on about four patches near the zenith, where it is easier to measure the brightnesses of celestial objects. Several factors complicate measurements near the horizon: the sky background is brighter because of the sun and nearby towns, celestial objects are dimmer because their light passes through more air to reach the telescopes, and the amount of air the light passes through changes rapidly with the elevation angle. Usually, RAPTOR monitors one patch of sky for several hours and then moves to another patch. While monitoring, RAPTOR takes two consecutive 30-second exposures through its A and B wide-field telescopes and analyzes the resulting digital images. If it sees an interesting event, RAPTOR zooms in for a closer look with the two narrow-field telescopes. The system also trains RAPTOR-S on the event to measure how the light intensity changes with wavelength. (RAPTOR-P, a very recent addition, looks for the slight dip in light intensity that can be seen from Earth when an extrasolar planet crosses its parent star's bright disk. This effect was used by a group at Princeton University to discover an extrasolar planet in 2001. RAPTOR-P also catalogs known celestial objects to help the system identify transient optical events.) [figure: RAPTOR observatories]

RAPTOR mimics human vision in its searches. If its computer brain detects something interesting in its wide-field "peripheral" vision, it quickly swivels its "eyes" to focus on the action. Then, like the cones in the fovea of the human eye, densely packed light sensors in RAPTOR's narrow-field central vision sharply image the region of interest. RAPTOR also uses two "eyes" for its stereovision.

Sky Monitoring

Weather permitting, RAPTOR-A and -B measure the positions and brightnesses of several million stars each night. On a moonless night, their wide-field telescopes can detect objects as faint as the 13 th magnitude. (By comparison, the faintest objects visible to the unaided eye—6th magnitude—are 300 times brighter.) Each narrow-field telescope has a field of view of 4 by 4 degrees and can detect objects as faint as the 17th magnitude. [figure: RAPTOR's fields of view]

Although RAPTOR is expected to find many transient optical events on its own or by responding to satellite alerts, the data collected through sky monitoring are valuable on their own. For example, by examining the data archives produced during ROTSE's sky monitoring, scientists were able to study the visible light accompanying a brief x-ray event months after a satellite detected it.

To Zoom or Not to Zoom?

To capture a transient optical event, RAPTOR must "think" and act fast. The system has one minute to decide if any one of up to 250,000 objects in a wide-field image is a transient optical event. In contrast, other robotic observatory systems capture transient optical events by chance or by responding to satellite alerts or to commands from humans, who are much slower and less precise than RAPTOR.

To detect a new object in a wide-field image, RAPTOR compares the position and brightness of each object in the image with those of known objects identified in previous scans. (Maintaining the list of known objects is itself a considerable computer task.) To make comparisons, RAPTOR first corrects for changes in the objects' apparent positions caused by optical aberrations in the telescope lenses and for changes in the objects' apparent brightnesses caused by atmospheric attenuation and lens vignetting (the loss of light near the edge of a lens). RAPTOR makes these corrections for all the objects in a wide-field image in 10 seconds or less.

When RAPTOR finds a new object in the first of the two consecutive images provided by both its A and B observatories, it determines if the object appears in the second images as well. This procedure eliminates artifacts produced, for example, by a cosmic ray passing through a light-sensor element in one of the digital cameras. False positives are a major problem for robotic observatories.

If an object passes this first test, RAPTOR measures the object's parallax to determine if the object is truly distant or merely an airplane, a meteor, a satellite, or a piece of space junk orbiting Earth. To qualify as a transient optical event, the object must be at least as far away as the moon.

RAPTOR measures parallax by comparing the two images from its A and B observatories. The positions of distant objects are the same in the two images, but the positions of close objects are different. This difference is the parallax. As with human vision, the closer the object, the greater the parallax. Stereo comparison also identifies malfunctioning light-sensor elements in the digital cameras, which can also mimic transient optical events.

No Shortage of Brains

To perform its complex tasks quickly, RAPTOR uses sophisticated software residing on nearly twenty personal and server computers. RAPTOR has far more computer intelligence than any other robotic observatory system.

Each telescope platform is controlled by its own personal computer. Also, a dedicated computer transfers the images from each digital camera to the system, which takes about 5 seconds per image. Because a computer is assigned to each camera, the transfers occur in parallel to reduce the total image-transfer time. For example, the total transfer time for a wide-field image, which consists of four slightly overlapping smaller images stitched together, is also about 5 seconds. The computers assigned to the cameras also measure the celestial objects' positions and brightnesses.

An image-server computer at each site analyzes data from the site's digital cameras and their computers for possible transient optical events. The image server then sends a list of candidate events to an alert-server computer located between the two sites. The alert server compares the lists from RAPTOR-A and -B observatories and identifies new objects that have no measurable parallax.

RAPTOR's master control consists of three computer programs that control the computers that in turn control the telescope platforms and the digital cameras. On the basis of data it receives from weather stations at each site, one of these programs also controls the observatories' domes. The programs also reside on a computer located between the two observatory sites.

Finding a Needle in a Haystack

RAPTOR's sky-monitoring data archives will provide valuable information. They offer a nearly continuous record of the light emitted by a large fraction of the celestial objects visible from Earth with a brightness of at least the 13th magnitude. Each week, sky monitoring adds up to 1 terabyte (1,000 gigabytes) to the data archives, which can be mined for interesting astronomical objects.

For example, a team of scientists from Los Alamos and the University of Michigan has used a computer program to identify 1,781 variable stars in the ROTSE data archives. (Variable stars help astronomers measure distances to other galaxies and study the galaxies' evolution.) About 90 percent of the variable stars had not been identified before. The computer program found the variable stars using the same process human astronomers follow. However, a team headed by Los Alamos astrophysicist Przemyslaw Wozniak has also programmed a computer devise its own ways to find patterns in the ROTSE data. This method could discover entirely new types of stars.

Wozniak's "machine-learning" technique resembles those used by credit card companies to detect fraud from anomalous spending patterns. (In fact, Los Alamos scientists developed some of the machine-learning techniques currently used by MasterCard to detect fraud.) In both cases, the techniques look for a few interesting patterns or events hidden in huge masses of data. In essence, they are looking for a needle in a haystack.

Defense scientists are also interested in finding "needles." They want to distinguish between the warheads and the decoys deployed by a hostile intercontinental ballistic missile, a scenario that includes many fast-moving objects with varying light intensities against a backdrop of stars. Defense scientists also want to detect the unique electromagnetic signature of a missile launch within a "forest" of signals whose sources range from cell phones to the sun. The techniques developed to help RAPTOR find needles in the cosmic haystack could help solve these problems as well.