Using Game Theory to Manage Self-Aware Unmanned Aerial Systems

Abstract

Runtime platforms on unmanned aerial systems (UAS) manage flight, GPS and compute resources and speed up common tasks for UAS workloads. These workloads evolve rapidly due to programmer demand and changing external conditions. Platforms are quickly out dated. Existing self-aware techniques update resource management policies and/or software, but managing the cost of updates under a budget is challenging. This paper makes the case for using game theory. Our approach profiles currently hosted workloads and measures efficiency gains from updates. Counter-factual regret, a game theory technique, computes when platforms should update. We outline a frame-work, provide an example and discuss research challenges.