An Online Mechanism for Resource Allocation in Networks

We consider the problem of designing an online mechanism for trading divisible resources over a constrained network. In an online setting, participating agents learn about their marginal valuations gradually over time. We propose a mechanism that implements the efficient allocation of resources with a finite-time guarantee, i.e., upon stopping after T > 0 rounds, the incentive to misreport information has a uniform bound that is O(1/root T). We illustrate the application of the proposed mechanism to optimize the efficiency in MHA networks, an emerging type of heterogeneous networks where users can simultaneously connect to several access technologies.