Profit-aware online crowdsensing task assignment for intelligent transportation services

With the rapid advancement of sensing and wireless technologies, intelligent transportation services (ITSs) have significantly enhanced the quality of public travel by capturing real-time traffic conditions. In practice, ITS platforms face bottlenecks in comprehensively acquiring urban traffic dynamics due to occasional sensor failures and coverage deficiencies, which, in turn, affect the quality of ITSs. To address these issues, crowdsensing, as an emerging computing paradigm, assigns traffic status sensing tasks (e.g., taking photos) to vehicle-based mobile participants (a.k.a. workers), improving the timeliness and spatial coverage of traffic monitoring. The above process raises a hot topic, i.e., online crowdsensing task assignment. Most existing methods primarily consider travel costs as the basic factor for pricing models to incentivize workers to complete crowdsensing tasks. However, due to the neglect of supply-demand dynamics in task pricing, these methods still suffer from imbalanced distributions of workers and tasks. In this paper, we propose a novel profit-aware online crowdsensing task assignment (POCTA) problem, which aims to maximize overall revenue by incorporating a supply-demand-aware pricing model. This model dynamically adjusts task prices based on current and predicted supply-demand conditions. We develop an efficient two-stage framework, predict-then-assign, to solve the POCTA problem. In the prediction stage, we build the end-to-end multi-view spatio-temporal attention network to predict the distributions of future crowdsensing tasks. In the matching stage, we propose the break-and-rematch online task assignment algorithm, which iteratively invokes a packing-aware matching operator and an adaptive assignment-breaking operator to optimize task assignment. Extensive experiments on two real-world datasets validate the efficiency and effectiveness of our proposed solutions.