OURS: optimal unicast routing systems in non-cooperative wireless networks

We propose novel solutions for unicast routing in wireless networks consisted of selfish terminals: in order to alleviate the inevitable over-payment problem (and thus economic inefficiency) of the VCG (Vickrey-Clark-Groves) mechanism, we design a mechanism that results in Nash equilibria rather than the traditional strate-gyproofness (using weakly dominant strategy). In addition, we systematically study the unicast routing system in which both the relay terminals and the service requestor (either the source or the destination nodes or both) could be selfish. To the best of our knowledge, this is the first paper that presents social efficient unicast routing systems with proved performance guarantee. Thus, we call the proposed systems: Optimal Unicast Routing Systems (OURS).Our main contributions of OURS are as follows. (1) For the principal model where the service requestor is not selfish, we propose a mechanism that provably creates incentives for intermediate terminals to cooperate in forwarding packets for others. Our mechanism substantially reduces the overpayment by using Nash equilibrium solutions as opposed to strategyproof solutions. We then study a more realistic case where the service requestor can act selfishly. (2) We first show that if we insist on the requirement of strategyproofness for the relay terminals, then no system can guarantee that the central authority can retrieve at least 1overn of the total payment. (3) We then present a strategyproof unicast system that collects 1over2n of the total payment, which is thus asymptotically optimum. (4) By only requiring Nash Equilibrium solutions, we propose a system that creates incentives for the service requestor and intermediate terminals to correctly follow the prescribed protocol. More importantly, the central authority can retrieve at least half the total payment. We verify the economic efficiency of our systems through simulations that are based on very realistic terminal distributions.