Device-to-device relay assisted cellular networks with token-based incentives

We consider a device-to-device (D2D) relay-assisted cellular network where mobile transceivers that are owned by self-interested users are incentivized to relay each other´s data using tokens, which they exchange electronically to “buy” and “sell” downlink relay services. We formulate the decision problem faced by each UE, namely, the problem of deciding whether or not to relay, as a Markov decision process (MDP). We propose a supervised learning algorithm that devices can deploy to learn their optimal relay policies online given their experienced network environment. Our simulation results show that, within the proposed token system, self-interested devices can achieve almost 15% higher throughput on average, and almost 40% higher throughput at the 90th percentile, than with only direct base-station-to-device communications. Additionally, we show that the token system performs best when the network contains neither too few nor too many tokens.