Optimizing Task Distribution Systems: A Comparative Study of Micro-Task Job Replication, Accuracy, and Budget Constraints

Crowd-sourcing harnesses the collective strength of a large and geographically diverse pool of individuals to tackle complex industrial tasks that surpass the capabilities of machine intelligence alone. This distributed approach is facilitated through crowd-sourcing platforms like Amazon Mechanical Turk (MTurk), CrowdFlower, Taskcn, and TopCoder. By leveraging the potential of crowd participation, these platforms enable the completion of tasks with significant monetary rewards, effectively reducing overall production costs for industries. In our work, we present a standardized model for industrial crowd-sourcing, where complex tasks are decomposed into manageable micro-tasks. Enthusiastic participants perform these micro-tasks while adhering to industry-defined accuracy standards within allocated budgets. Our crowd-sourcing platform follows a reactive execution process, involving planning, assignment, and execution stages. Additionally, we introduce a task monitoring algorithm that provides performance control mechanisms, allowing for strategic budget utilization and establishing a robust foundation for the platform. Simulation reveals five copies outperforming three copies in achieving $A$*, with three copies requiring 76 times more tasks. Five copies exhibit 82% success vs. three copies' 77%, reducing task incompleteness. Increasing task copies challenges budget constraints, and machine accuracy $\alpha_{M}=0.5$ achieves 18.12% and 4.825% more completion at a higher cost.