Zhilin Xu

This thesis examines strategic decision-making in settings where individual incentives may conflict with collective efficiency. Motivated by this tension, it studies the emergence and stability of cooperation under competition, incomplete information and changing participation. The analysis considers coalition formation in Bertrand model and strategic interaction in Mobile Crowdsensing systems with heterogeneous requesters and participants. Despite the differences between these settings, the thesis addresses a common economic question: under what conditions can cooperation be sustained? Chapter 2 studies cooperation in competitive markets. Using a Bertrand model with collusion, it examines the effects of the number of firms on coalition formation and characterizes the size of stable coalitions under different market sizes. The results show that collusion can persist in equilibrium as the number of firms increases. The analysis also shows how the formation of coalitions affects prices and profits in the market. Chapter 3 develops a cooperation-driven model for the MCS system with multiple heterogeneous requesters. By modeling the interactions between requesters and participants as multiple sub-games, the proposed model enables collaborative data collection while maintaining individual rationality of the requesters. Based on this model, a payment method and a cost-sharing method are designed to improve data collection quality and ensure the benefits of cooperation. To address the case of incomplete information, a parameter learning method based on reinforcement learning (Q-learning) is further introduced to estimate participants’ private information from historical data, thereby supporting the design of dynamic payment strategies. Chapter 4 develops a dynamic model to characterize the MCS system where requesters can join or leave the system while also demands can change over time. In this setting, a staged matching algorithm and corresponding state-update rules are designed to dynamically coordinate requester demands with participant resources. The proposed approach reduces the system cost while improving frequency of requester participation and maintaining the stability of the matching outcome. Chapter 5 further addresses the challenges arising from incomplete information and the irregular presence of participants by proposing a participant-centric matching mechanism. The mechanism performs participant-centric task allocation to adapt to the dynamic changes in participants’ presence. Furthermore, a requester classification algorithm is designed to reduce the computational complexity of the matching algorithm. Theoretical analysis demonstrates that the proposed mechanism achieves stable matching performance across all stages of the decision-making process. Chapter 6 addresses the time-sensitive nature of sensing tasks by proposing a dynamic coopetition mechanism that integrates competition and cooperation. In the competition stage, a budget-constrained matching algorithm is designed to improve the utility of requesters. In the cooperation stage, a data redistribution strategy is introduced to enhance the overall task completion rate of the system. Theoretical analysis and simulation results demonstrate that, compared with a purely competitive mechanism, the proposed dynamic coopetition mechanism can significantly improve the overall task completion rate while maintaining an acceptable utility level for individual requesters. Taken together, this thesis examines how cooperation can arise and persist under different forms of strategic interaction. The first part focuses on coalition formation in Bertrand competition, while the second part studies incentive mechanisms for cooperation in Mobile Crowdsensing systems. Despite their different settings, both parts examine how individual incentives and collective outcomes are connected.