ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2022, Vol. 59 ›› Issue (2): 310-328.doi: 10.7544/issn1000-1239.20210875

所属专题: 2022空间数据智能专题

• 人工智能 • 上一篇    下一篇

一种基于时空位置预测的空间众包任务分配方法

徐天承1,乔少杰1,武俊2,韩楠3,岳昆4,易玉根5,黄发良6,元昌安7   

  1. 1(成都信息工程大学软件工程学院 成都 610225);2(西南财经大学证券与财经学院 成都 610074);3(成都信息工程大学管理学院 成都 610225);4(云南大学信息学院 昆明 650500);5(江西师范大学软件学院 南昌 330022);6(南宁师范大学计算机与信息工程学院 南宁 530023);7(广西教育学院 南宁 530023) (ticxu2019@qq.com)
  • 出版日期: 2022-02-01
  • 基金资助: 
    国家自然科学基金项目(61772091,61802035,61962006,61962038,U1802271,U2001212,62072311);四川省科技计划项目(2021JDJQ0021,22ZDYF2680,2021YZD0009,2021ZYD0033);成都市技术创新研发项目(2021-YF05-00491-SN);成都市重大科技创新项目(2021-YF08-00156-GX);四川音乐学院数字媒体艺术四川省重点实验室资助项目(21DMAKL02);CCF-华为数据库创新研究计划项目(CCF-HuaweiDBIR2020004A);广西自然科学基金项目(2018GXNSFDA138005);成都市“揭榜挂帅”科技项目(2021-JB00-00025-GX);四川省科技创新苗子工程项目(2021006)

A Spatial Crowdsourcing Task Assignment Approach Based on Spatio-Temporal Location Prediction

Xu Tiancheng1, Qiao Shaojie1, Wu Jun2, Han Nan3, Yue Kun4, Yi Yugen5, Huang Faliang6, Yuan Chang’an7   

  1. 1(School of Software Engineering, Chengdu University of Information Technology, Chengdu 610225);2(School of Securities and Futures, Southwestern University of Finance and Economics, Chengdu 610074);3(School of Management, Chengdu University of Information Technology, Chengdu 610225);4(School of Information Science and Engineering, Yunnan University, Kunming 650500);5(School of Software, Jiangxi Normal University, Nanchang 330022);6(School of Computer and Information Engineering, Nanning Normal University, Nanning 530023);7(Guangxi College of Education, Nanning 530023)
  • Online: 2022-02-01
  • Supported by: 
    This work was supported by the National Natural Science Foundation of China (61772091, 61802035, 61962006, 61962038, U1802271, U2001212, 62072311), Sichuan Science and Technology Program (2021JDJQ0021, 22ZDYF2680, 2021YZD0009, 2021ZYD0033), Chengdu Technology Innovation and Research and Development Project (2021-YF05-00491-SN), Chengdu Major Science and Technology Innovation Project (2021-YF08-00156-GX), Digital Media Art, Key Laboratory of Sichuan Province, Sichuan Conservatory of Music, Chengdu, China (21DMAKL02), CCF-Huawei Database System Innovation Research Plan (CCF-HuaweiDBIR2020004A), Natural Science Foundation of Guangxi (2018GXNSFDA138005), Chengdu “Take the lead” Science and Technology Project (2021-JB00-00025-GX), and Science and Technology Innovation Seedling Project of Sichuan Province (2021006).

摘要: 空间众包技术在现实物理世界中有着丰富的应用场景,得到学术界和工业界的广泛关注.任务分配是空间众包的主要研究问题之一,即把工人分配给合适的任务.但是现有的任务分配方法大多假设众包工人和空间任务出现的位置和时间是已知的,忽略了真实的众包平台中众包工人和空间任务的动态变化,由于空间众包平台的强时效性,这种情况下设计的分配方式只能得到局部最优分配结果.提出最大价值最小成本任务分配的新问题,目标是对当前和未来的工人进行分配,使用最小的移动成本获得最大的分配价值.为解决这一问题,提出了基于轨迹的任务分布预测方法及基于核密度估计的工人分布预测方法,设计基于位置预测的任务分配算法来计算众包工人和空间任务的相对最优分配策略.所提位置预测方法利用图卷积神经网络和ConvLSTM模型进行预测,相较传统基于网格的位置分布预测更加精确和稳定.基于位置预测的启发式分配算法可以在线性时间内结合预测得到的位置信息完成任务分配,更加契合空间众包平台的强时效性.在真实数据集上进行大量实验来证明所提方法的有效性,相比于基于网格的预测方法,任务/工人位置预测准确率分别提高了15.7%和18.8%.

关键词: 空间众包, 在线任务分配, 空间数据智能, 位置预测, 最小成本

Abstract: Space crowdsourcing technology has a varying type of application scenarios in the real physical world, which has been widely concerned by academia and industry. Task assignment is one of the important research issues in space crowdsourcing, that is, distributing workers to appropriate tasks. However, most existing task assignment methods assume that the location and time of crowdsourcing workers and space tasks are known, ignoring the dynamic change of crowdsourcing workers and space tasks in real crowdsourcing platforms. Due to the requirement of strong timeliness in space crowdsourcing platforms, in this situation, the designed assignment method can only get the local optimal results. A new problem of task assignment with maximum value and the minimum cost is proposed, which aims to distribute current and future workers and obtains the maximum assignment value by using the minimum traveling cost. To handle this problem, a trajectory based model is proposed to predict the distribution of tasks. In addition, a kernel density estimation based model is proposed to predict the distribution of workers. A task allocation algorithm based on location prediction is designed to calculate the relative optimal assignment strategy between crowdsourcing workers and spatial tasks. The proposed location prediction method uses graph convolution neural network and ConvLSTM model to predict, which is more accurate and stable than the traditional grid-based location distribution prediction approaches. The heuristic assignment algorithm based on location prediction can complete task allocation in a linear time by combining the predicted location information, which is consistent with the strong timeliness of the space crowdsourcing platform. Extensive experiments are conducted on real datasets to prove the effectiveness of the proposed methods. Compared with the grid-based prediction method, the accuracy of task/worker location prediction is increased by 15.7% and 18.8%, respectively.

Key words: spatial crowdsourcing, online task assignment, spatial data intelligence, location prediction, minimum cost

中图分类号: