基于蒙特卡罗贝叶斯推理的移动群智感知可靠任务分配机制  

Reliable task allocation mechanism in mobile crowd sensing based on Monte Carlo Bayesian inference

作　　者：杨桂松[1] 姚秋言 Yang Guisong;Yao Qiuyan(School of Optical-Electrical&Computer Engineering,University of Shanghai for Science&Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机应用研究》2022年第11期3365-3370,3384,共7页Application Research of Computers

基　　金：国家自然科学基金资助项目(61802257)。

摘　　要：针对现有任务分配策略的不足,研究了在工人数量有限的移动群智感知系统中任务分配策略,借助社交网络来分配任务并获得高收益。首先,建立了社交网络的动态不确定环境,利用社交网络完成任务,传播任务。然后考虑到不同社交网络对任务的偏好不同,设置任务偏好度这一不确定指标,借助经济学风险价值的理论描述任务分配的可靠性。最后利用蒙特卡罗贝叶斯推理方法研究任务动态传播模型的复杂参数的高斯过程,设计基于知识梯度的采样算法选择蒙特卡罗采样点,从而实现高收益的任务分配方案。为了验证所提策略的性能,将其与四种基准的采样算法进行比较。实验结果表明,所提任务分配策略在提高收益方面是有效的。In view of the shortcomings of existing task allocation strategies,this paper proposed a task allocation strategy in mobile crowd sensing with a limited number of workers and used social networks to allocate tasks to obtain high profits.Firstly,it established the dynamic uncertain environment of social network to complete and propagate task.Then,considering that different social networks have different preferences for tasks,this paper set the uncertain value of task preference,and described the reliability of task allocation with the help of the theory of economic value at risk.Finally,it used Monte Carlo Bayesian inference method to study the Gaussian process of complex parameters of task dynamic propagation model,and designed a sampling algorithm based on knowledge gradient to select Monte Carlo sampling points,so as to realize high profit task allocation strategy.In order to verify the performance of the proposed strategy,it is compared with four benchmark sampling algorithms.The experimental results show that the proposed task allocation strategy is effective in improving profit.

关 键 词：移动群智感知 任务分配 风险价值 动态不确定 蒙特卡罗贝叶斯推理 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]