面向多目标群智感知的任务分配算法研究

Crowdsensing makes use of a great number of participants equipped with mobile sensing devices to implement various large-scale sensing applications that penetrate into people’s life. As the core algorithms of crowdsensing, sensing task assignment algorithms have become one of the research hotspots in the research community. Most of the existing research works on crowdsensing focus on optimizing sensing tasks from the same category. These works construct single objective models, based on which task assignment and data collection methods are designed. They have not incorporated the spatiotemporal relation features of different kinds of tasks, and thus cannot conduct collaborative optimization of tasks belonging to multiple categories. To overcome this problem, this proposal aims to investigate multi-objective task optimization models for crowdsensing with different kinds of tasks. By analyzing the factors, such as uncertainty of participant trajectories and dynamic features of participants, that influence the performance of algorithms for multi-objective models of sensing tasks, we will propose assignment approaches based on trajectory estimation, coverage estimation, and dynamic interaction models. The proposed assignment algorithms can make full use of the multi-dimensional sensing capabilities of participants, and thus improve the efficiency of task assignment algorithms for crowdsensing. Finally, we will verify the performance of the designed task assignment algorithms on the sensing applications in crowdsensing-based intelligent transportation systems. The results of this project will improve the overall performance of task assignment for crowdsensing and promote the real-world applications of crowdsensing systems.

群智感知利用携带移动感知设备的大量参与者来实现大规模的、渗透到人们生活方方面面的各类感知应用。作为群智感知的核心算法，感知任务分配算法已成为该研究领域的研究热点。目前大部分群智感知研究专注于单一类型的感知任务，构建单目标优化模型并进行任务分配和数据收集利用，未能融合不同类型任务之间的时空关联特性并进行多类型任务的协同优化。针对该问题，本项目研究面向多类型任务的多目标群智感知优化模型，针对多目标感知任务模型中参与者轨迹不确定性和动态特性等影响算法性能的因素，分别提出基于轨迹估计、覆盖估计、以及动态交互的分配策略，充分利用感知参与者的多维度感知能力，提高群智感知任务分配算法的效率。最后在基于群智感知的智能交通感知应用中对以上新型任务分配算法进行应用检验。本项目研究成果将提高群智感知任务分配的整体性能，促进其在实际系统中的应用普及。

感知任务分配是群智感知的一个核心问题，目前大部分分配算法未能融合不同类型任务之间的时空关联特性并进行多类型任务的协同优化。针对该问题，本项目研究多目标群智感知的任务优化模型，基于用户轨迹、时空覆盖、动态交互等任务分配中的关键元素，系统地研究了群智感知任务分配算法理论、机制、方法、以及系统。具体的，针对主动式感知和机会式感知场景的特征，研究了一系列基于用户轨迹规划、用户轨迹模糊估计、用户感知时长估计、位置熟悉度估计等模型的高效分配算法。针对感知任务不同粒度的覆盖需求，研究了基于跨区域动态预算、低预算任务聚类、双边隐私保护等技术的用户激励和任务匹配算法。针对感知系统用户动态交互的挑战，研究了基于空间多样性、任务完成质量、平台效用优化、感知用户异构等元素的群智感知激励与分配算法。项目相关成果已发表高水平国际期刊和会议论文30篇，包括IEEE/ACM Transactions系列论文13篇，授权发明专利6项。这些成果提高了群智感知任务分配和执行的整体效率，促进了群智感知系统的应用普及。

内容获取失败，请点击重试