Distributed Strategic Learning for Effective Gossiping in Wireless Networks

Novel applications running over wireless networks are currently emerging, and concrete examples are Vehicle Ad Hoc Networks, Internet of Things or Multimedia Streaming. Such applications are characterized by multicast communication patterns, and a serious vulnerability to packet loss patterns, which are severe when radio frequency signals are used for communicating. Therefore, reliable multicasting over wireless networks is demanding to be properly resolved. Gossiping is a widely known and successful approach to reliable communications, but reliability is achieved at the costs of worse performances and a heavy traffic load on the network, in terms of an increased number of exchanged messages to obtain a successful delivery. Our driving idea to improve the efficiency of gossiping is to make use of a distributed strategic learning for determining the best set of nodes to send a gossip message, so as to optimize the utility of such messages. We have experimentally assessed such a solution through a preliminary set of simulations.