【作者】 周华强；

【导师】 江海峰；

【作者基本信息】 中国矿业大学 ， 软件工程（专业学位）， 2021， 硕士

【摘要】 移动群智感知具有成本低廉、覆盖范围广和维护简单等特点,不同于传统静态传感器网络,它利用用户身上携带的移动设备作为感知单元,实现感知任务的分发和感知数据的收集。目前,在环境及交通状况监测等领域,移动群智感知有着广泛的应用前景。在移动群智感知中,任务分配策略和用户的激励机制是两大关键研究问题。在现有研究中,针对噪声监测这类时空跨度广的任务分配策略研究较少,大时空跨度场景下带来的用户疲劳度差异大,时空复杂度问题尚未得到有效解决。同时,在预算有限的前提下,采用拍卖等报酬方式进行激励机制设计存在部分困难任务得不到有效招标和任务执行质量得不到保证的问题。本文针对城市噪声监测中时空覆盖广、不同时段用户疲劳度差异大和数据收集质量低等特点,设计了基于时空联合优化的移动群智感知任务分配策略。针对移动群智感知中用户参与积极性低、困难任务无人认领和任务完成质量得不到保证等情况,设计了基于工程招投标和拍卖结合的移动群智感知双流程激励机制。(1)基于时空联合优化的移动群智感知任务分配策略。移动群智感知中,噪声监测场景的任务分配效率低,其大空间跨度会增加任务分配复杂度,大时间跨度会导致参与用户疲劳度差异大,影响任务执行效果。该策略采用集合覆盖思想对城市噪声监测这类大时空任务分配问题进行建模,综合考虑大时空跨度下的用户疲劳度差异和任务参与者历史任务完成情况,以任务时空限制作为约束条件,构建任务分配性价比指标作为优化目标,采用启发式的遗传算法进行问题求解。(2)基于工程招投标和拍卖模型结合的移动群智感知双流程激励机制。现有的基于报酬的激励机制未考虑感知用户的任务执行能力,用户对困难任务的参与积极性不高。本文所提的双流程激励机制按时间先后包括总体任务招投标环节和剩余困难任务拍卖环节。总体任务招投标环节将用户的投标报价和感知能力相结合进行综合评价,构建平台效益指标,选择平台效益得分最高的用户完成任务招标。针对剩余无人投标的困难任务,设计了提高预算利用率的办法,通过引入改进的荷兰式拍卖机制,进一步提升用户参与的积极性和整体激励的成功率。基于MATLAB平台,本文分别测试了上述任务分配策略和激励机制的性能并与相关算法进行对比分析。结果表明,本文提出的任务分配算法在任务完成质量、性价比指数和任务执行冗余度等方面性能表现良好,平均性能提升幅度约26.5%;所提出的双流程激励机制在预算利用率、平台总效益和激励成功率方面与同类算法相比性能较好,平均性能提升幅度约17.6%。该论文有图23幅,表11个,参考文献64篇。更多还原

【Abstract】 Mobile crowd sensing has the characteristics of low cost,wide coverage and simple maintenance.Unlike traditional static sensor networks,mobile crowd sensing uses the mobile device carried by the user as the sensing unit to realize the sensing task distribution and collection of perception data.At present,in the fields of environment and traffic condition monitoring,mobile crowd sensing has a wide range of application prospects.In mobile crowd sensing,task assignment strategies and user incentive mechanisms are two key research issues.In the existing research,there are few researches on task assignment strategies with wide temporal and spatial spans such as noise monitoring.The user fatigue caused by large temporal and spatial span scenarios is very different,and the problem of temporal and spatial complexity has not been effectively solved.At the same time,under the premise of limited budget,the use of auctions and other compensation methods for incentive mechanism design has the problems that some difficult tasks cannot be effectively tendered and the quality of task execution cannot be guaranteed.Aiming at the characteristics of wide temporal and spatial coverage of urban noise monitoring,large differences in user fatigue in different periods,and low data collection quality,this paper designs a mobile crowd sensing task assignment strategy based on space-time joint optimization.Aiming at the situation of low user enthusiasm for participation in mobile crowd sensing,unclaimed difficult tasks,and unguaranteed task completion quality,a dual-process incentive mechanism for mobile crowd sensing based on the combination of engineering bidding and auction is designed.(1)Task assignment strategy for mobile crowd sensing based on space-time joint optimization.The task assignment efficiency for noise monitoring scenarios with a large time span is low.A large space span will increase the complexity of task assignment,and a large time span will lead to a large difference in the fatigue of participating users and affect the effect of task execution.This strategy uses the collective coverage idea to model large-scale spatio-temporal task assignment problems such as urban noise monitoring,comprehensively considers the differences in user fatigue under the largescale spatio-temporal span and the historical task completion of task participants,and takes the task time-space constraints as constraints to construct,the task assignment cost performance index is taken as the optimization target,and the heuristic genetic algorithm is used to solve the problem.(2)Dual-process mobile crowd sensing incentive mechanism based on engineering bidding and auction models.Existing reward-based incentive mechanisms do not consider perceiving the user’s task execution ability,and the user’s enthusiasm for participating in difficult tasks is not high.The dual-process incentive mechanism mentioned in this article includes the overall task bidding link and the remaining difficult task auction link according to time.The overall task bidding link combines the user’s bidding quotation and perception ability for comprehensive evaluation,constructs platform benefit indicators,and selects the user with the highest platform benefit score to complete the task bidding.In response to the remaining difficult tasks of unmanned bidding,a method to improve budget utilization was designed,and an improved Dutch auction mechanism was introduced to further enhance the enthusiasm of users to participate and the success rate of overall incentives.Based on the MATLAB platform,this paper tests the performance of the abovementioned task assignment strategy and incentive mechanism and compares and analyzes them with related algorithms.The results show that the task assignment algorithm proposed in this paper shows good performance in terms of task completion quality,cost performance index,and task execution redundancy,with an average performance increase of 26.5%.The proposed dual-process incentive mechanism is effective in budget utilization,total platform benefits,and incentive success.Compared with similar algorithms,the performance is better in terms of rate,with an average performance increase of 17.6%.更多还原