SCOLDING software could speed the hunt for Earth-like planets by punishing robotic telescopes that don鈥檛 perform as well as promised.
Last week, astronomers reported spotting the first multiplanet system orbiting another star, thanks to a network of robotic telescopes dotted around the world (see 鈥淥ne-hit-wonder planet-spotting technique adds a new feather to its cap鈥�). Each telescope runs software that allows it to alert the entire network as soon as it spots something interesting. The telescopes then bid for the chance to carry out follow-up observations, citing their assessment of their equipment, position and availability.
Not all astronomers like automated telescope networks however. The networks can make mistakes that humans wouldn鈥檛, leaving astronomers with useless results, says Alasdair Allan at Exeter University in the UK. For instance, a telescope that is not connected to a weather station might offer to make observations even when it is cloudy, he says. Others overestimate the quality of their equipment relative to other telescopes in the network.
Advertisement
To avoid such mistakes, Allan and his colleagues have built a (www.arxiv.org/abs/0802.0431) that ranks telescopes on the data they return. If a telescope fails to keep its promises, the system scolds it, instructing it to work out why it failed and to be more modest. 鈥淚f the telescope repeatedly doesn鈥檛 listen and continues to enter bids that promise too much, the software will punish it by ignoring it,鈥� Allan says. Because it is programmed to modify its behaviour to maximise bids, the telescope gradually learns to be more modest.
The team hopes to install the software on the network, which links telescopes in Australia, Hawaii and La Palma. George Djorgovski at the California Institute of Technology in Pasadena, says that such software will be essential as more telescopes go robotic. 鈥淭he new software can make sophisticated decisions.鈥�
