Emerging Directions in Aerospace Software V&V
Jim Barhorst, James L. Paunicka, Douglas A. Stuart, Jonathan A. Hoffman · Infotech@Aerospace 2011 · 2011
Approaches to developing safety-critical aerospace software have been remarkably successful, especially in the case for piloted aircraft, owing largely to disciplined software development practices and effective interaction between developers and customer / regulatory stakeholders. However, currently emerging trends in vehicle software size and complexity, including autonomy, will benefit from new work in verification, validation, and certification (VV&C). This is echoed by a number of commentators on the research landscape, including the National Research Council, a number of other government agencies such as NASA, the DoD and, more recently, the Chief Scientist of the United States Air Force (USAF). The National Research Council has suggested that the current methods of code review and testing be augmented by analysis techniques for generating evidence of high levels of dependability for increasingly complex and autonomous systems. Several of the top eight prioritized research focus areas in the USAF Chief Scientist’s May, 2010 publiclyreleased “Technology Horizons” report of key areas of science and technology explicitly include the development of novel verification and validation (V&V) methods (especially for emerging autonomous systems). Apart from their application to additional vehicle capabilities and autonomy, improvements in VV&C methods will benefit the development of emerging flight systems that exhibit similar complexity to today’s flight platforms, thereby making such platforms more affordable and enhancing industrial competitiveness by reducing the high cost of current methods for VV&C of flight systems. Boeing is seeking new, cost-effective approaches yielding comparable or enhanced levels of assurance, while accommodating advanced adaptive and autonomous capabilities required to satisfy future needs. Initial efforts have focused on new VV&C technologies and capabilities that can be incorporated into existing processes. This includes both traditional and the newer agile and lean processes that are making their way into the aerospace domain. We have particularly emphasized early analysis, automated testing, and automation of the VV&C process for model-based development. These initial steps should provide significant near- to mid-term benefits. We are also defining the requirements for more advanced approaches. Three areas we have identified for further study are compositional VV&C, online/runtime VV&C, and science to support VV&C of adaptive and autonomous systems. Compositional system development is enabling cost and cycle reductions, but compositional VV&C approaches that provide comparable benefits are only beginning to emerge. With the emergence of unmanned platforms designed for extremely long endurance (e.g., multiple days, years), with concomitant needs for in-flight software configuration changes for both fault management and capability enhancement, approaches for performing VV&C at “runtime” (in flight) may be required. Thirdly, existing VV&C approaches are geared towards systems that are both static and deterministic, while future autonomous and adaptive platforms and ultra large scale cyber-physical systems of systems will be increasingly dynamic and will appear increasingly non-deterministic. This will require new V&V approaches, and the science to 1