BioEd Online

European scientists have sent a 'cyborg' to roam the Spanish countryside as part of a mission to create robots that are good at exploring planets independently.

Researchers at the Centre for Astrobiology near Madrid kitted out a human with a camcorder linked to a computer system programmed to look for interesting features in the landscape. The human merely did the donkey-work of carrying the hardware while the computer did the 'thinking'. On a planetary mission, a robotic vehicle such as NASA's rovers Spirit and Opportunity, currently touring the surface of Mars, would carry the hardware.

Initial field tests of this cyborg astrobiologist showed that the computer-controlled vision system could identify some of the same geological features that human geologists would have selected as being worthy of closer study. The computer selects its targets using an 'uncommon map', which is an analysis of the features in an image that least resemble the rest of the picture.

Proponents of human space exploration often argue that robots are no match for trained astronauts and geologists in spotting promising study sites and responding to chance discoveries. But if the current work fulfils its promise, future robotic explorers will have a decision-making capacity similar to that of human experts.

Remote control

Robotic explorers would be useful, not just on Mars, but also on more remote and inclement worlds, such as Jupiter's icy moon Europa and Saturn's haze-shrouded moon Titan, which NASA's Huygens spacecraft is due to explore early next year.

The Spirit and Opportunity rovers are controlled from Earth by programming them at the beginning of each martian day with all the moves they must execute before they shut down to conserve power during the night. The vehicles have some degree of autonomy, however. For instance, they might be programmed to go to a particular location, but in getting there they are free to execute spontaneous manoeuvres to avoid obstacles.

Even so, this makes the rovers very single-minded. If they were to encounter an interesting site en route, or to spot a destination that looks more promising that the programmed one, they would not be able to act on their own initiative. The astrobiologist programme aims to develop a system that has this kind of flexibility and 'intelligence'.

Significant difference

The system's mapping software, developed by the Madrid team and computer scientists at the University of Bielefeld, Germany, mimics the behaviour of real geologists scanning a new scene. They tend to pay most attention to areas that stand out as different from the rest. The hardware is all wearable by a single person, and the researchers tested it at a cliff face in Madrid's Southeast Regional Park. Their results will be published in the International Journal of Astrobiology¹.

To find truly 'interesting' areas on the cliff face, the cyborg system has to overcome various challenges. For example, it has to be able to distinguish shadows from areas of genuinely different rock colour. It has to be able to 'remember' what it found initially interesting as it zooms in on a certain candidate site; a region that stands out at a distance might look blandly uniform close up. And it has to decide what kind of differences to look for: colour, texture or light intensity, for instance.

The researchers admit that some of these problems can still fox their system. Team leader Patrick McGuire says that they have yet to develop an intelligent zoom facility for the camera, for example. They are encouraged, however, that on one field trip the system identified precisely the spot on the cliff face that the geologists in the team agreed was the most promising, where water had seeped out of the rock to darken its surface. Spotting a feature like this would be particularly important in the search for water on Mars.

Model army

McGuire says that other groups are working on similar 'robot geologists', for example at NASA's Ames Research Center in Mountain View, California. "They may be a little more advanced," he admits, but he points out that his "baby geologist" has the virtue of being unbiased: it simply looks for stand-out features rather than being preprogrammed to find, say, a specific type of mineral.

The Madrid team's system is now scheduled to explore Spain's Rio Tinto area, where the highly acidic and iron-rich ground water is thought to mimic some of the extreme conditions on other planets. In collaboration with NASA, scientists at the Centre for Astrobiology are planning to drill a hole several metres deep and to send a worm-like robot to seek out interesting features in the underground rock.