Strategic Availability and Cost Effective UAV-based Flying Access Networks: S-Modular Game Analysis
For UAV operators in temporary wireless coverage scenarios, this provides a game-theoretic framework to optimize pricing and availability under energy constraints.
This paper analyzes a duopoly game between UAV service providers competing on pricing and availability, showing the game is sub-modular in availability and super-modular in pricing, and proposes a best-response dynamics learning scheme for distributed strategy learning.
Telecommunication service providers deploy UAVs to provide flying network access in remote rural areas, disaster-affected areas or massive-attended events (sport venues, festivals, etc.) where full set-up to provide temporary wireless coverage would be very expensive. Of course, a UAV is battery-powered which means limited energy budget for both mobility aspect and communication aspect. An efficient solution is to allow UAVs switching their radio modules to sleep mode in order to extend battery lifetime. This results in temporary unavailability of communication feature. Within such a situation, the ultimate deal for a UAV operator is to provide a cost effective service with acceptable availability. This would allow to meet some target Quality of Service while having a good market share granting satisfactory benefits. We construct a duopoly model to capture the adversarial behavior of service providers in terms of their pricing policies and their respective availability probabilities. Optimal periodic beaconing (small messages advertising existence of a UAV) is a vital issue that needs to be addressed, given the UAVs limited battery capacity and their recharging constraints. A full analysis of the game outcome, both in terms of equilibrium pricing and equilibrium availability, is derived. We show that the availability-pricing game exhibits some nice features as it is sub-modular with respect to the availability policy, whereas it is super-modular with respect to the service fee. Furthermore, we implement a learning scheme using best-response dynamics that allows operators to learn their joint pricing-availability strategies in a fast, accurate and distributed fashion.