Incentive Allocation to Sequential Decision-Making Sensors in Mobile Crowdsensing

In this work in progress, we consider incentive allocation to a set of measurement sensors in the context of mobile crowdsensing. The novelty stems from considering a new model perspective for each sensor, that of a rational sequential decision-maker. At each time slot, each sensor observes the time-varying cost it undergoes for submitting measurements and the advertised reward for submitting measurements to the platform. Its decision policy at each time slot is whether to become active and submit measurements or stay inactive. The sensor decision problem is shown to be described as an optimal stopping one, and the sensor policy that maximizes its expected net benefit over a time horizon is shown to be of threshold nature at each time slot, where the threshold is non-increasing with the elapsed time. With the derived optimal policies for sensors, we next seek to determine the optimal price per time slot paid by the platform to each sensor so as to maximize the expected total quality of collected measurements, subject to a budget constraint. Finally, we introduce the problem of centralized sensor activation in a dynamically varying system so as to maximize the long-term average utility stemming from the quality of collected data. The characterization of distributed sensor equilibrium policies and the assessment of their impact on the global performance metric compared to the optimal centralized policy, are outlined as important directions that warrant further investigation.