The unprecedented growth of Internet of Things (IoT) enables smart devices to connect with each other, leading to a wide range of ubiquitous applications. In the Lot paradigm, crowdsourcing is considered to be a promising approach for providing efficient sensing, computing, and processing service to a particular task generated by customers, efficiently integrating the power of crowd. In this paper, we investigate the crowdsourcing platform utility maximization by finding the optimal pricing policy for requested tasks. Such a pricing strategy can be formulated as a finite sum optimization problem in which the nodes try to achieve a global consensus on the pricing policy of each task. During the process of optimization, however, some nodes may be in the sleeping mode so that they cannot perform instantaneous updates, and it is too time-and resource-consuming to proceed synchronously centralized optimization due to the large scale networks. To address this issue, we use the stochastic gradient descent (SGD) type algorithm nonconvex primal-dual splitting with exact minimization (NESTT-E) to solve the optimization problem distributedly and asynchronously. Numerical results show the NESTT-E is more efficient than synchronous ADMM and conventional SGD with a larger number of working nodes.
展开▼
