Incentivizing crowdsourcing for radio environment mapping with statistical interpolation

Abstract

White Space Networking crucially relies on the active monitoring of spectrum usage (to identify white space opportunities) in both space and time. One way to achieve this is wide-area deployment of spectrum sensors to gather spatio-temporal spectrum data, and use them to construct better Radio Environment Maps (REMs) via suitable statistical interpolation techniques (i.e., Kriging). Cost of such large-scale sensor deployment can be reduced via crowdsourcing, i.e., outsourcing the sensing task to mobile users equipped with sensorized high-end client devices (e.g., tablets or smartphones), and success of such crowdsourced sensing presumes some incentive mechanisms to attract user participation. In this work, we present an incentivized crowdsourcing system architecture that (periodically) acquires spectrum data from users, so as to optimize the resulting radio environment map (i.e., minimizing the average prediction-error variance) for a given data acquisition budget. First, we introduce an auction-based incentive mechanism that is computationally efficient, individually rational and truthful, and prove that the total payment of the proposed mechanism is a monotonically increasing function of the cardinality of the winner set. Then we propose a budget-feasible version and through extensive simulations, we evaluate the performance of proposed mechanisms for comparison to a baseline to demonstrate its significantly superior performance in crowdsourced radio mapping.