FUTURE missions to Mars and the moon will use robotic rovers that essentially think for themselves. Built using sophisticated hardware and software, the rovers will be able to decide when they are in trouble, how to solve the problem, and take steps to protect themselves against catastrophe.
The principle has already been proven by the Mars rover Spirit, which NASA spectacularly revived on the surface of the Red Planet earlier this year. In what NASA calls a 鈥減rescient design decision鈥� it equipped Spirit and its twin, Opportunity, with the capability to diagnose and respond to certain faults. It was a decision that saved Spirit from an ignominious end.
The successful rescue means that intelligent fault-protection technology is set for far wider use. First up will be a NASA rover called the Mars Science Laboratory, planned for launch in 2009. It will attempt a hazardous 20-kilometre drive across the planet鈥檚 surface over two years, but will be equipped with sensors and artificial-intelligence software to detect and evaluate risky terrain or conditions. Different operating modes will allow mission controllers to access and adjust the rover鈥檚 computer memory.
Advertisement
The European Space Agency, too, may harness self-preservation technology in a proposed Mars sample-return rover mission scheduled for 2011. Also any rovers at the vanguard of new US missions to the moon and Mars, announced by President Bush in January, will be equipped in this way.
Self-preservation software will be critical. It will take the ESA rover, for instance, at least 18 minutes to send a radio signal to Earth and receive instructions back from mission control. That鈥檚 too long if it runs into trouble. When it does, the rover must be able to take action to preserve three things: communications links are the top priority, followed by power and thermal management to ensure the craft is not crippled by the 鈭�90 掳C nights. 鈥淵ou cannot fly this kind of mission without a fault-protection system that is very, very deep,鈥� says Gentry Lee, a systems engineer at NASA鈥檚 Jet Propulsion Laboratory (JPL) who has worked on most planetary missions since the mid-1960s.
Just 18 days after Spirit landed in Gusev crater on 2 January, the craft was crippled by a software glitch. This caused problems with the 256 megabytes of flash memory attached to the craft鈥檚 computer, sending it into a rapid, progressive decline that caused it first to cease communicating and then to communicate only in odd bursts that were largely gibberish. Managers were alarmed, initially thinking that at best the craft would be seriously disabled, and at worst that it might be defunct.
But a variety of mechanisms allowed it to preserve some power in its batteries, and send enough data home to give controllers a peek at the problem. The forward-thinking design of the craft鈥檚 fault-protection system, Lee says, made it possible 鈥渢o get in there to do some diagnostic work while it was sick but not dead鈥�.
After two weeks of detective work and testing on a mock-up, the problem was traced to the way the software handled large numbers of files stored in the flash memory. It was then fixed by reformatting the memory and carefully monitoring the number of files that built up in future.
Spirit鈥檚 recovery was not only a relief for the scientists, it was an important demonstration that it is possible to build complex computerised missions 鈥� as Bush鈥檚 plan will demand. In future it will not be feasible to plod through the old NASA process of trying to predict every eventuality and test responses. And with software growing ever more complex as it grows more capable, the ability for it to self-diagnose a problem and take steps to protect the craft until mission controllers can find out what has gone wrong is crucial.
鈥淚 can鈥檛 name one spacecraft that could have survived what Spirit went through,鈥� says Lee. Three months later, in late April, the rover team finally reached the end of the recovery process: a set of software patches was beamed up over a three-day period and successfully installed. These included fixes to ensure that the flash memory problem will not recur on Spirit and will never happen on Opportunity.
But perhaps even more importantly, the experience has shown designers of future planetary rovers what the vehicles need to be able to do to stay alive and wait for help when things go wrong.
Heel, rover
While engineers at JPL were diagnosing rover problems on Mars, their colleagues at NASA鈥檚 Ames lab in Silicon Valley have been taking a different tack. They are planning ways to use computerised rovers to boost the productivity of people on Mars, if they ever get there.
The Ames team is developing a wheeled robotic rover, called the Extra-Vehicular Activity Robotic Assistant (ERA), that can automatically follow and monitor astronauts on field trips. ERA will take on many of the same chores that mission controllers performed for the Apollo astronauts on the moon: keeping track of their schedules, tasks accomplished and their remaining power and oxygen, for example. The idea is to leave the astronauts free to collect materials and measurements, and make observations, without having to keep track of the logistics.
The Ames team, led by Bill Clancey, is now approaching the end of two weeks of field research in the Utah desert, in which researchers in spacesuits have been conducting geological fieldwork, accompanied by a four-wheeled rover that follows voice commands from its designated master astronaut.
