Jan Norstrom developed a new method to improve software design and reduce system failures. This new method, based on decision analysis, could improve the success rate of satellite missions in space.
Risk assessment is a major topic in a wide variety of disciplines today, covering fields from financial risk assessment on the stock exchange to risk assessment for safety and reliability in engineering and design. Although the areas of application differ, the same mathematical rules apply, addressing major issues like uncertainty in terms of probabilities and decision analysis.
These are the same conceptual principles used in control software designed by Jan Norstrom, a 29 year old Norwegian, who conducted his PhD project at the Department of Control Risk, Optimisation, Stochastics and System Theory.
Norstrøm successfully developed a method to guide the design of control software architecture via decision theory. From a strong theoretical and mathematical basis, the new method can be used to improve the architectural design of software systems for automated satellite operations.
Norstrøm's thesis focused on the control software program that should activate the antenna boom deployment mechanism after launch and separation. During a launch four satellites were mounted together. Attempting to deploy antennas during launch or before separation could result in the loss of the satellites. If the actuators are activated too late, the satellites might not reach the correct orbit, or a recovery mission would have to be undertaken to try to fix the problem.
Losing scientific satellites is a very expensive failure: The loss of the Ariane 5.01 rocket, which exploded with four Cluster satellites onboard worth about 500 million Euro, was caused by errors in the booster control software.
Prior to graduating, Norstrom worked at the European Space Research and Technology Centre (ESTEC), in Noordwijk, where he worked with validation methods for control software called Sneak Path Analysis. His work at ESTEC convinced him that there was a definite need for methods that could quantitatively evaluate the architectural design of software.
When Norstrom asked why architectural design was done in a particular way, he received a stock response: The current architecture works; so why change it? Norstrom was not satisfied with this reply: "If optimisation of the control software‘s decision-making process can utilise sensor information slightly better, there’s a potential to reduce risk, improve reliability and ensure the success of satellite missions in space. Even a small improvement in reliability can be worth millions of dollars."
Besides reducing the risk of failure in automated satellite operations, Norstrom's new method has many other application possibilities. This is owing to the mathematical basis of the decision theory in his work. In fault-diagnostic, the power lies in reducing the vast number of tests needed and finding failures in a more systematic manner. The technique can be applied when designing other forms of software architecture, such as in the medical field, where the process of diagnosing cancer, for example, can be simplified and made more cost-effective.
Another notable example of a software system that could be vastly improved by Norstrom's new method is Windows. "My hope is that new versions of Windows will use my technique, so it will stop crashing several times a day. But if Bill Gates is going to be the boss of the Windows company after Microsoft is broken up, I very much doubt it," Norstrom says, smiling.
Although Norstrom has returned to Norway, where he works for Kongsberg, a defence and aerospace company, he is not lost to science and will continue to do research in this field. As for his years in the Netherlands, Norstrom has very fond memories: "I miss my friends in Holland a lot. I loved living in the Netherlands and may come back someday. But in the Netherlands I missed the Norwegian nature and winter sports. I've been out training slalom every Thursday this winter."