Adaptive Data Offloading in Opportunistic Networks through an Actor-Critic Learning Method

Nowadays, the growing popularity of mobile phones has re- sulted in an exponential growth of mobile data, causing severe traffic overload in the cellular network. A promising approach to overcome this problem is offloading, i.e. to delegate part of the traffic to other networks. In this paper we consider offloading through opportunistic networks of users' devices. Clearly, this strongly depends on mobility patterns, therefore achieving efficient and timely content delivery could be very challenging. In this paper we propose an adaptive offloading solution based on an actor-critic algorithm, which is a type of reinforcement learning algorithm widely used in control problems. More precisely, in our solution the controller of the dissemination process, once trained, is able to perform at any time the most appropriate choice about the number of content replicas to be injected in the opportunistic network to guarantee the timely delivery of contents to all interested users. Our system is able to automatically learn the best strategy to reduce the traffic on the cellular network, without relying on any additional context information about the opportunistic network. Finally, our solution reaches higher level of offloading w.r.t. other state of art approaches, in a range of different mobility settings.