IE 11 is not supported. For an optimal experience visit our site on another browser.

Artificial astronomers work while you wait

From Space.com: A team of astronomers, seeking to streamline the capabilities of telescope observation, has developed a computer program that can watch the skies for them.
Nik Szymanek
/ Source: Space.com

A team of astronomers, seeking to streamline the capabilities of telescope observation, has developed a computer program that can watch the skies for them.

Used in conjunction with a robotic telescope and an online connection, the Intelligent Agent — as the program is called — can sift through sky observations and even compare them with prior images of the same region. The program can then decide whether it has found something interesting enough for further study and makes more observations if needed.

While the Intelligent Agent lacks the computing sophistication — and homicidal tendencies — of the fictional "Matrix" program Agent Smith, it should help astronomers track dramatic sky events like supernova explosions while the researchers themselves are elsewhere. If such an event is detected, the agent could simply page or call its human controllers.

"It's the lack of human intervention that's important here," explained astronomer Andy Adamson, head of the Joint Astronomy Center/United Kingdom Infrared Telescope, or UKIRT, in Hilo, Hawaii, where the new agent made its debut. "It's really a clever kind of system."

Intelligence in the machine
The viability of using computerized agents to fill in for flesh-and-blood researchers hinges on two basic technologies, the online access and robotic telescopes.

Robotic telescopes themselves are nothing new. Research into dates back to the 1960s at least, with a host of new robotic skywatching projects currently in the works. Online databases collected by Internet-ready telescopes have led to completely that allow researchers to make new discoveries sifting through archived data.

But to build a truly autonomous telescope requires at least some measure of intelligence to make observation decisions. To do this, a group of British researchers formed the eScience Telescopes for Astronomical Research project at the UKIRT. The project is better-known by its acronym, eSTAR.

After developing the Intelligent Agent software and testing it out on a network of small telescopes, the eSTAR team directed the program to make live observations with the 12.4-foot (3.8-meter) UKIRT instrument. During that larger test, the agent compared UKIRT images with data of the same region by the James Clerk Maxwell Telescope, also run by the Joint Astronomy Center in Hawaii, and discovered a dwarf nova star, an object that can go through sudden flares in brightness.

"We've now proved that any telescope system which has a queue of observations waiting to be done can be accessed by eSTAR," explained eSTAR project leader Tim Naylor in an e-mail interview. UKIRT, he added, is not completely robotic, but eSTAR's success with the instrument means the system can be used with non-automated as well as automated systems. "That just increased the usefulness of this system tenfold," Naylor said.

The only human intervention involved in the UKIRT-JCMT test was from UKIRT controllers accepting the Intelligent Agent's observation request. But eSTAR researchers said the programs still have much to learn about astronomical research.

"The current agents are fairly primitive, but the architecture allows you to extend them, [meaning] they can get smarter," Alasdair Allan, an eSTAR project member and physics professor at the University of Exeter, explained during an e-mail interview. "Eventually, I anticipate that we'll be using neural networks and genetic algorithms to allow them to learn for themselves." Such advancements, Allan added, would allow computer more autonomy when it comes to decision-making and cut down on the amount of time an agent would contact human controllers for guidance.

Telescopes that phone homeCommunication between computer agents and human astronomers is a critical component of the eSTAR project. Once the agent confirms detection of a target object, it needs a way to alert astronomers of the stellar find.

Gamma ray bursts, for example, are some of the most powerful explosions in the universe, but last from a fraction of a second to minutes at most. Researchers hoping to study these objects can either stare at one portion of the sky and hope one occurs while they're watching, or tap into the GRB Coordinates Network that collects burst data from a number of orbiting satellites and sends out alerts via the Internet, pagers and e-mail.

But that system focuses on gamma ray bursts, while eSTAR's Intelligent Agent could be set to report any event thought to be critical by its human user.

"The universe is a quirky sort of place, you can’t turn your back on it [because] it changes too fast," Allan said. "Eventually we anticipate that the astronomers will be able to interact with their agents using their 3G mobile phones, allowing them to be in constant contact with the software that's carrying out their science program."

Allan expects agent programs to serve UKIRT astronomers next spring during a scheduled research program on gamma ray bursts. The agents will be able to send text message alerts to that project's researchers, he added.

The eSTAR team is also working to create a larger network of telescopes for the agents to use. In addition to UKIRT, which is not robotic, project researchers plan to include the James Clerk Maxwell Telescope and fully robotic telescopes such as the Liverpool Telescope, on La Palma in the Canary Islands, and the two Faulkes Telescopes in Hawaii and Australia.

A research GridInstead of designing Intelligent Agents to roam the Internet, the online arena used by most people to send e-mail, images and other data, eSTAR researchers decided a more scientific medium was in order. The Intelligent Agents use a system called the Grid, also known as the "next-generation Internet."

The key difference between the Internet and the Grid comes down to access. The Internet grew to be a tool typically accessed by humans, while the Grid is being developed as a resource for distributed processing. The project, which uses computers around the world to sift radio telescope data for signs of intelligent extraterrestrial life, is one example of a distributed processing.

"It isn't so much the hardware as the philosophy behind it," Allan said. "The Grid is being built from the ground up so that it's components are interoperable, [whereas] the Internet just grew, which isn't a bad thing, just different."

The software interfaces of Grid services are well-specified and designed to be accessed by other software like the Intelligent Agent program, which expected to move back forth between a telescope control system and the observation databases of other instruments, Allan added.

Keeping romance in the researchWhile certainly useful, Intelligent Agents can also be seen as one more buffer between human astronomers and the sky they desperately seek to observe. The advent of modern detectors and cameras, after all, has consistently pushed astronomers from the telescope eyepiece to the computer mainframe, where they have more control and data at their fingertips.

"I actually think [Intelligent Agents] add a bit more excitement to the field," Adamson said, adding that astronomers often receive gamma ray burst data long after the burst actually occurred. Intelligent agents could reduce the delay between detecting a burst and observing it from 10 hours to about 30 minutes, he added.

While Adamson sees the Intelligent Agent system as a boon for flesh and blood astronomers, those same living, breathing researchers can themselves be an obstacle for the system.

"The main challenge we face now are organizational and sociological," Naylor said. For example, will observatories accept Intelligent Agents requesting telescope time rather than human astronomers? And if so, would astronomers really give computerized counterparts a high degree of autonomy? "For those who do, I'm sure there's a wealth of discoveries waiting to be made," Naylor said.