Employing Stochastic Multiplayer Games to Support Self-Organization over Ad Hoc Networks

Abstract

Self-organization over ad hoc networks is of growing interest due to their application to collective adaptive systems of IoT devices, such as wearables and drones. These devices can possess situational goals that depend on the availability of external services provided by service providers (SPs) within the ad hoc network. SPs have quality-of-service (QoS) attributes whose values are based on the service(s) they have available. These QoS attributes can be negatively affected by environmental sources of uncertainty as well as behavioral constraints imposed on the SP to support self-organization via integration. Novel mechanisms are needed to evaluate the impact of an integration configuration on the SP's QoS attributes to produce Pareto optimal configurations. We construct a stochastic multiplayer game (SMG) that evaluates a SP's expected satisficing level given its privileged data access, sources of uncertainty, and QoS values. Polynomial regression is applied to the output of the SMG to produce a model to evaluate an integration configuration at runtime. We demonstrate the model on a rescue scenario involving wearables and drones and examine the efficacy of the resulting configurations.