To err is human… Can computers fix our mistakes?

Developing software that automatically detects errors

On June 4, 1996, the Ariane 5 launcher exploded less than a minute after take-off. The accident was not the result of a mechanical fault but rather due to an error in the design of the guidance software. At the Department of IT a research team is developing software to automatically detect and circumvent mistakes of this type.

Today it is possible to completely design and verify complex computer chips before the first prototype has been built. First engineers design the chip using specialised software, then a computer can simulate this chip and automatically find weak points with the help of mathematical methods.

These developments have been rapid. The first Pentium processors made mistakes when they divided one number by another. Today most makers of computer chips use software to discover and correct design glitches before the product goes to market. Since chips are getting more and more complex, researchers are forced to steadily improve their methods for making ever faster verification programs. A further area where automatic verification is useful is communications protocols, such as those used in mobile telephones to make it possible for people to communicate with each other. The first generation of mobile phones was limited to voice transfer, but modern equipment can transfer images and video films. New protocols are needed. Every protocol has to be able to guarantee that data is received by the proper destination within a reasonable period of time.

There are many other applications as well. The fact that computers are making their way into more and more systems means that the field of new uses is constantly expanding. The need to develop new algorithms is growing apace.

Johann Deneux & Pritha Mahata