Joint Optimization of System and User Oriented Task Assignment in Mobile Crowdsensing

One of the fundamental challenges in mobile crowdsensing (MCS) systems is efficient task assignment. Existing solutions consider the problem from system's point of view and try to maximize the system utility or minimize the cost of sensing. However, such a task assignment process does not consider the user (i.e., workers and task requesters) preferences and can yield unhappy users with their assignments, impairing the participation of users in the future. To incentivize the user participation, stable matching based solutions can be utilized to result in satisfactory assignments that will make the users happy based on their preferences. However, this may adversely affect the system utility especially when the set of eligible number of workers for each task is limited. To address this issue, we study the task assignment problem in MCS systems that maximizes the main system utility (i.e., number of workers and tasks assigned) as a system oriented goal while generating as happy users as possible with their assignment. As the problem is NP-complete, we first solve the problem optimally using Integer Linear Programming (ILP) and then we propose a heuristic based polynomial solution that runs very fast. Through simulations, we show that the proposed approach achieves the maximum possible system utility while generating small and close to optimal user unhappiness as in ILP results.