Incentive mechanisms for non-proprietary vehicles in vehicular crowdsensing with budget constraints

Abstract

Vehicular crowdsensing (VCS) utilizes the onboard sensors and computational capabilities of smart vehicles to collect data across diverse regions. Non-dedicated vehicles, due to their lower cost and broad distribution, have emerged as a central focus in VCS research. However, their trajectories are often concentrated in urban areas, resulting in uneven data coverage. Existing incentive mechanisms primarily rely on platforms to dynamically adjust task allocation based on vehicle trajectory predictions. Yet, they frequently neglect the influence of geographic locations on vehicle routing choices and fail to incentivize proactive route planning. To address this, we propose a novel two-phase incentive mechanism that, for the first time, incorporates a willingness to traverse factor. This mechanism aims to maximize spatial coverage within a limited budget by encouraging vehicles to voluntarily traverse remote areas to complete tasks. In the initial phase, a multi-agent deep reinforcement learning algorithm dynamically adjusts each vehicle’s route and quote price, which is then reported to the platform. In the second phase, the platform allocates tasks and adjusts compensation based on the provided routes and quotes to optimize overall platform benefits. Experimental results show that our mechanism effectively balances platform and vehicle benefits, achieving optimal outcomes even under budget constraints.