Evaluating Collaborative and Autonomous Agents in Data-Stream-Supported Coordination of Mobile Crowdsourcing

Area of Science:

• Computer Science
• Artificial Intelligence
• Operations Research

Background:

• Mobile crowdsourcing systems rely on an on-demand workforce for tasks requiring physical movement.
• Current systems often assign tasks to unsuitable crowdworkers, leading to high failure rates and diminished service quality.

Purpose of the Study:

• To develop and evaluate mechanisms for predicting task outcomes and coordinating task transfers in mobile crowdsourcing.
• To address challenges posed by worker autonomy and uncertain task outcomes in dynamic environments.

Main Methods:

• Analysis of various data stream learning approaches for accurate task outcome prediction.
• Development and evaluation of two distinct task coordination strategies: an opportunistic approach for crowdshipping and a market-based model for crowdsensing.

Main Results:

• Demonstrated the effectiveness of data stream learning for predicting task success in mobile crowdsourcing.
• Showcased the viability of proposed coordination mechanisms in improving task completion and service quality under different worker autonomy levels.

Conclusions:

• The proposed prediction and coordination mechanisms offer a robust solution to mitigate failure rates in mobile crowdsourcing.
• These approaches enhance service quality by enabling adaptive task reassignment, crucial for both crowdshipping and crowdsensing applications.