Integrating autonomous fault management with conventional flight software: a case study

Abstract

An evolutionary approach to on-board autonomy involves integrating autonomous processes with traditional spacecraft real-time flight software (FSW). Such an approach relies on layering autonomy onto existing architecture while making minimal changes to the existing FSW. This is distinctly different from a revolutionary approach, which calls for integrating the design of autonomy and FSW from the start. While a revolutionary approach may lead to a more optimal design, an evolutionary approach allows for reuse of proven FSW, facilitates the complicated process of software verification and validation, and accelerates acceptance by the aerospace community. Because intelligent autonomous agents are typically soft real-time processes while FSW typically meets hard real-time requirements, applying the evolutionary approach is not a straightforward task. One technique available for solving this problem is Rate Monotonic Analysis (RMA). RMA uses real-time scheduling theory to guarantee that all processes can be scheduled. It has been used successfully in real-time scheduling applications such as helicopter control systems. This paper presents a case study where RMA was used as a tool to resolve real-time scheduling and processing issues associated with adding autonomous fault isolation and recovery software to previously developed flight software. The combined software system was then successfully tested on a real-time testbed.