Improving the quality of service indices of task allocation in mobile crowd sensing with fuzzy-based inverse stackelberg game theory

•This paper presents a novel approach using fuzzy logic and deep reinforcement learning for mobile crowd sensing task allocation, achieving over 17% higher coverage quality compared to traditional methods.•The proposed method combines game theory and fuzzy logic to optimize resource allocation, reducing average service delay and maximizing profit for all actors in the Internet of Things and fog computing context.•To address real-world challenges, the paper utilizes neural networks for approximation in deep reinforcement learning, ensuring stability, convergence, and adaptability in mobile crowd sensing task allocation. This paper introduces a comprehensive strategy for heterogeneously allocating tasks, aiming to optimize mobile crowd sensing through the use of fuzzy logic and thus achieving superior coverage quality. We employed a deep learning method to address the diverse range of requests. Recognizing the instability during the learning process, we utilized an approximation function for the Q-values, thereby preventing divergence during the training phase of the model. A prominent challenge is ensuring robust user participation in mobile crowd sensing initiatives. Essentially, a higher number of monitoring nodes within an area correlates with improved coverage quality. We employed fuzzy logic to estimate participation density, taking into account both the duration of users' presence in the study region and the geographical density. Our results are compelling: the proposed method boosts coverage levels by over 17% compared to standard techniques. Additionally, with an accuracy spanning 91.5% to 95.3% for the correct allocation of resources using a dataset from Google, the efficacy of our approach is further underscored.