A Dynamic Assurance Framework for an Autonomous Survey Drone

Typical practice for software safety assurance requires the generation of large amounts of assurance data, which can be complex and very expensive to maintain. This assurance data is often presented in the form of an assurance case or safety case, which justifies that the software is considered acceptably safe for use in a given context. Many modern systems are also difficult to assure without being very conservative about worst case performance, particularly when using technology such as multi-core processors and GPU accelerators. This is exacerbated when they are deployed in dynamically changing environments, and means resources can be under utilised. In this paper we present a framework for dynamic risk assessment and assurance of a highly configurable autonomous unmanned drone. We show how the use of continually updated confidence metrics, combined with dialectic arguments, can support a more agile dynamic assurance case approach. We examine two example monitors in detail, and link a battery charge monitor to the assurance case using a dialectic approach to communicate the impact and meaning of confidence shortfalls. We comment on our findings and link to other related work.