Brains in Space: Why Your Toaster is Smarter than NASA

Brains in Space: Why Your Toaster is Smarter than NASA

Adam Frank for McGraw-Hill

Space is a dangerous place to have a brain, at least a silicon one. For all their glamour, interplanetary probes have never been very smart. Remember the intrepid pathfinder probe on Mars? Most people assumed it landed with the equivalent of a supercomputer on board. The truth is far less exotic. There are smarter hand-held calculators around today than what NASA was able to put on the Mars probe. To survive the dangers of deep space, scientists and engineers always have been forced to dumb-down the silicon intelligence used to guide their NASA missions. It's a trade-off that has left spacecraft with stilted abilities far below what exists in even the average desktop PC. With such limited capabilities, scientists have faced difficult choices in terms of what data to collect and how much of it to save. It's a situation that may have worked fine for the first 40 years of space exploration but fails miserably in the new millennium. NASA now is contemplating robotic Martian rovers traveling on their own for tens of miles each day or deep-sea probes hunting for signs of life in the (possible) oceans of Jupiter's Moon Europa. These missions demand probes that can operate on their own without waiting for commands to arrive from the distant earth (it could take 20 minutes or more for signals to reach a Europa probe). NASA wants the era of truly autonomous spacecraft to arrive fast. That will not be possible without really smart spacecraft.

What's the Problem?
The space environment is a bad place to put electronic circuits. Outside of Earth's protective blanket of air there waits a constant bombardment of dangerous high-speed charged particles. The Sun is not kind to interplanetary spacecraft. Dangerous energetic particles from the Sun interact with matter in the space probe including on-board microprocessors. Tiny but deadly, these subatomic bullets can tear through a craft and wreak havoc on delicate electronic equipment. Circuits are all about conducting electricity. The charged particles from the Sun are powerful enough to dive deep into the guts of a probe changing the patterns of charge on the internal chips. The deposition of charged particles on microchips can stop their functioning altogether or produce errors in their calculations. The most insidious of these errors are called "bit flips," simple changes of 0's to 1's in a microprocessor's memory. Scrambling the spacecraft's brains this way can mean a loss of valuable data that the robot explorer traveled a long way at great expense to gather. It's enough to give any NASA engineer goose bumps.

Radiation hardening has been the traditional solution to the problem. Engineers can create new chips by changing the microchip design (the layout of circuits on the chip). They also can change the actual material the chip is made from or change the manufacturing process. These two strategies allow them to create microprocessors relatively immune to the effects of radiation damage. Unfortunately, creating radiation-hardened chips is expensive and time consuming. Redesigning a circuit means changing delicate timing patterns within its thousands of tiny components. The redesigns also mean the latest tricks of the trade cannot be incorporated into the radiation hardened chips. The clever insights incorporated into the Pentium III Processor that give it so much speed will not be on the spacecraft's chip. In the end, the microprocessors space scientists and engineers have at their disposal are brawny in terms of withstanding hardships but they are slower, more expensive and have higher power needs that what is available in the commercial market. In terms of hard data, radiation hardened processors are often three to 10 years behind the state-of-the-art and also can be 10 times more expense as well.

Given their super-stupidity, lack of speed and the energy-hog power needs, no one is going to create a genius Martian robot rover with rad-hardened microprocessors. What is the solution? There are a number of ideas floating around but one of the most direct has been to go ahead and use off-the-shelf commercial chips on the spacecraft but use software to detect when an error has occurred. There are sophisticated ways in which, via software programs, a chip can police itself to check if radiation has caused any problems in its own workings. This idea looks quite promising for correcting errors like bit flips and other problems that don't lead to the chip being completely fried by radiation. It won't work for so-called mission critical systems where you need to make sure the microprocessor works in all conditions.

Space is a dangerous place. If we are truly to make it a home we still have a lot to learn.

Points to Ponder

The dangers from radiation in space change with location and time. Why do you think this might be true?
Can you think of some locations and particular times in the solar system when a spacecraft will be in the most danger?
Can you think of two specific reasons why NASA wants autonomous spacecraft?