Incentivizing Crowdsourcing Workers via Social Networks in Edge Computing
- 11 December 2024
- journal article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Consumer Electronics
- Vol. 71 (2) , 5653-5665
Abstract
Crowdsourcing has proven to be an effective approach for leveraging collective intelligence to solve tasks. However, in edge computing environments, crowdsourcing platforms often face challenges such as insufficient worker participation and weak collaboration. Existing incentive mechanisms focus on individual contributions, which leads to inefficient collaboration and suboptimal resource utilization, particularly on resource-constrained edge devices. Furthermore, limited computational capacity and budget constraints hinder widespread participation, negatively impacting task performance and overall engagement. This paper introduces the Agent Cooperation Model, which enhances worker collaboration and task completion through a distributed, agent-based approach to task allocation. We also propose the Deep Reinforcement Learning-based Dynamic Worker Incentive Mechanism (DDWM) to optimize incentive distribution using social network data, ensuring more effective incentives and fostering greater participation and collaboration. Cooperation Optimization through Multi-Agent Reinforcement Learning (MARL), allowing ACM to dynamically improve resource allocation and cooperation efficiency. The expansion of high-influence worker incentives through DDWM, driving broader participation via social networks. Experiments conducted on real-world datasets confirm that ACM and DDWM consistently enhance worker collaboration and participation across diverse and complex social networks. Compared to other incentive methods, these models exhibit superior stability and convergence, particularly in larger and denser networks, validating the effectiveness of our approach in fostering efficient crowdsourcing cooperation.Keywords
Funding Information
- Fundamental Research Funds for the Central Universities (36330603)
- National Natural Science Foundation of China (61672122, 62002039)
This publication has 34 references indexed in Scilit:
- A deep reinforcement learning based long-term recommender systemKnowledge-Based Systems, 2021
- A social mechanism for task-oriented crowdsourcing recommendationsDecision Support Systems, 2021
- A parallel team formation approach using crowd intelligence from social networkComputers in Human Behavior, 2019
- Incentivizing Long-Term Engagement Under Limited Budget2019
- Adaptive Incentive Allocation for Influence-Aware Proactive Recommendation2019
- Crowdsourcing Software Task Assignment Method for Collaborative DevelopmentIEEE Access, 2019
- A Context-Aware Multiarmed Bandit Incentive Mechanism for Mobile Crowd Sensing SystemsIEEE Internet of Things Journal, 2019
- Online Incentive Mechanism for Mobile Crowdsourcing Based on Two-Tiered Social Crowdsourcing Architecture2018
- Incentivize Multi-Class Crowd Labeling Under Budget ConstraintIEEE Journal on Selected Areas in Communications, 2017
- Measuring the social influences of scientist groups based on multiple types of collaboration relationsInformation Processing & Management, 2017