Quasi-Optimal Dual-Phase Scheduling for Pigeon Networks

In situations where communication infrastructures are destroyed or do not exist at all, implementing a disruption/delay-tolerant network (DTN) may be the only choice. A special type of DTN, which is known as a pigeon network, utilizes controllable special-purpose vehicles called pigeons to convey messages among segregated nodes and gateways. This paper studies a basic scenario in a pigeon network, where the segregated nodes are in an area that is far from the gateway. Our focus is on the optimal scheduling problem that determines how long a pigeon should stay at each node so that the average delay of messages is minimized. Although this problem is related to the server vacation model, existing literature in this area mainly focuses on the performance evaluation of different scheduling schemes, whereas the study on the optimal scheduling schemes is rarely seen, particularly for bursty traffic. In search of the optimal scheduling scheme, we first introduce a peak-valley message generation model and investigate the optimal scheduling schemes for the peak phase and the valley phase, respectively. Based on the analytical result, we then propose a quasi-optimal dual-phase (QODP) scheme. We further design a robust phase identification scheme such that a pigeon is able to apply the locally optimal scheduling scheme at the corresponding phase. Simulation studies have demonstrated QODP's adaptability in facing different types of traffic and its significant performance improvement over the existing scheduling schemes.