A Lyapunov Optimization Approach for Long-Term Task Assignment in Vehicular Crowdsensing

Vehicular crowdsensing has recently received considerable attention, due to its promising capability of collecting useful information for the Internet of Vehicles. However, existing researches in crowdsensing mainly focus on optimizing the short-term objectives, e.g., quality of crowdsensing, social welfare, or crowdsensing cost. Few of them consider the assignment of crowdsensing tasks from a long-term perspective, particularly for tasks located within a sensing hole. Due to the lack of sufficient participants nearby, these tasks may not be timely completed and have to be delayed. Particularly, in a dynamic crowdsensing system, the requirements of tasks may also change over time, which presents a new challenge in achieving the system stability. To tackle these issues, this paper formulates a novel task assignment problem from a long term perspective. A Lyapunov optimization model is constructed to optimize both the stability of the system and the utility of crowdsensing. Additionally, we establish the upper bound of the utility function, based on which we construct an auxiliary graph to convert the defined task assignment problem into finding a minimum weight maximum flow from the graph. Finally, we perform simulations to validate the efficiency of the proposed scheme.