Value of information based design of control software

This paper presents a suggested alternative to simplistic majority voting schemes based on the value of uncertain information. It uses satellite antennae deployment as an illustrative example. Control software is used in satellites to activate system functions like, e.g. antenna deployment. The software receives observations from sensors built into satellites and uses this information to trigger required functions. Often, inadvertent activation and delayed response can have severe consequences. Hence the way in which sensor information is processed strongly influences the system performance. We discuss an approach that models various ‘design’ options in detail so that the software control flow can be optimised via decision theory. We give some mathematical background and an example based on the CLUSTER satellite system that was spun off a design problem at European Space Research and Technology Centre (ESTEC). Our example considers the decision of when to deploy a satellite antenna. The control software must decide when to inspect sensors and when to deploy the antenna. We show how to optimise both the inspection time and the time to deploy the antenna given the results of the inspection. For our example it is important that the consequences of the control software decisions are analysed and measured in monetary loss associated with failure. This allows us to measure the risk in expected loss of money. Given control software designs A and B one can compare them by obtaining the value of information.