Neural-Network Aided Dynamic Control for Delivering Media Streams in Selfish Wireless Networks With Unknown Node-Selfishness

In this paper, we investigate the effect of selfish behaviors on the media stream delivery in selfish wireless networks with selfish relay nodes (RN). The RN's selfish behavior of forwarding media streams, depicted as its degree of node-selfishness (DeNS), is affected by both its available resources and the received incentives, while an incentive mechanism is employed to depress the selfish behaviors of the RNs having limited available resources. Under the RN's node-selfishness, the optimal incentives are controlled by the sources delivering media streams to maximize the finite-horizon utility, namely, the tradeoff between the path reliability and the incentive cost for delivering media streams during a finite time period. With the unknown node-selfishness owing to the dynamic information about the RNs' available resources and their received incentives being private, we provide a dynamic node-selfishness model to formulate the dynamic of the RN's DeNS, and further design an online neural-networks aided approximation scheme to identify the node-selfishness dynamics and to determine the optimal incentives. The convergence of this scheme is analyzed by Lyapunov techniques. Simulation results demonstrate the effectiveness of delivering media streams.