Abstract
Mobile Crowd Sensing (MCS) leverages the ubiquitous presence of smartphones and wearables to create large-scale sensor networks. However, traditional MCS models often assume a single central requester, ignoring the competitive reality where multiple entities vie for the same pool of workers. This paper presents CENTURION, a framework that addresses multi-requester MCS by integrating a double-auction incentive mechanism with reliability-aware data aggregation. CENTURION ensures truthfulness, maximizes social welfare, and guarantees data accuracy through a computationally efficient greedy approximation algorithm.
Introduction
The proliferation of mobile devices has enabled Mobile Crowd Sensing (MCS) to become a powerful tool for environmental monitoring, traffic management, and urban sensing. Despite its potential, most existing MCS research focuses on a "one-to-many" model. In practice, multiple requesters—such as different municipal departments or private firms—often compete for the same crowd workers. Furthermore, the design of incentive mechanisms (to encourage participation) and data aggregation (to ensure quality) are frequently decoupled. CENTURION bridges this gap by creating a unified system that balances the strategic interests of all participants while maintaining strict data quality standards.
System Architecture
The CENTURION framework operates as a tripartite ecosystem consisting of requesters, a central platform, and workers. The platform acts as the intermediary, managing both the economic incentives and the technical data processing.
As shown in the architecture, the process follows a structured sequence:
- Sensing Requests & Bids: Requesters submit sensing tasks and their corresponding bids (willingness to pay).
- Task Allocation & Worker Bids: The platform presents task sets to workers, who then submit bids based on their execution costs.
- Incentive Mechanism (MELON): The platform selects a set of winners (requesters and workers) and determines payments.
- Data Submission: Winning workers provide data labels for their assigned tasks.
- Data Aggregation: The platform combines these labels into a final result for the requesters.
- Final Payments: Payments are processed to ensure fairness and truthfulness.
The CENTURION Framework
CENTURION is built on two primary pillars:
- The MELON Double Auction: Unlike standard auctions, MELON allows both requesters and workers to submit bids simultaneously. The platform selects winners to maximize social welfare while ensuring "truthfulness"—meaning the best strategy for any participant is to bid their true value or cost, preventing strategic gaming of the system.
- Weighted Data Aggregation: To handle noisy or conflicting sensor data, CENTURION moves beyond simple majority voting. It employs a weighted approach that prioritizes data from workers with higher historical reliability, shifting the final result toward the most trustworthy data points.
The core innovation lies in the coupling of these mechanisms. The winner selection process in the auction is mathematically constrained by the accuracy requirements of the aggregation phase. Since finding the optimal set of winners is NP-hard, the authors developed a greedy approximation algorithm that remains computationally efficient even as the number of participants scales.
Evaluation and Key Results
The framework was evaluated through extensive simulations involving up to 150 workers and 80 requesters. Key findings include:
- Social Welfare: CENTURION generated significantly higher social welfare compared to baseline models like MSW-Greedy and AIR.
- Accuracy: The weighted aggregation mechanism resulted in a much lower Mean Absolute Error (MAE) than standard mean or median aggregation methods.
- Reliability: The empirical error probability consistently remained below the user-defined threshold ($\beta_j$), validating the framework's theoretical guarantees.
- Efficiency: Despite the underlying complexity, the proposed greedy algorithm demonstrated polynomial-time complexity, making it viable for real-world MCS platforms with thousands of users.
Conclusion and Future Work
CENTURION provides a robust solution for the complexities of multi-requester MCS. By linking incentives directly to data quality, it creates a sustainable ecosystem for both data consumers and providers. Future research will focus on addressing the "cold-start" problem for new workers without historical reliability data, extending the framework to continuous data types (like temperature or GPS), and incorporating advanced privacy-preserving techniques like differential privacy into the auction process.
Summary updated