获取物理世界的信息是物联网所提供的各种应用和服务的基础。当前对于传感器感知问题的研究大部分是假设理想的部署环境或充足的传感器数量，而事实上物联网的很多应用中，传感器的感知能力、部署成本以及物理环境等因素造成物联网难以全面感知监测区域，传感器的部署和数据采集常常受到限制。针对传感器部署和数据采集受限条件下物联网感知能力提高进行研究，可以扩展物联网的应用范围，并成为近年来物联网领域研究的热点。 本文以提高物联网在传感器部署和采集受限条件下优化物联网感知能力为目标，全面而系统地研究了物联网的传感器受限的部署优化和数据采集的一些关键问题。主体内容分为两个部分：传感器优化部署策略和数据采集关键技术。前者主要研究事件感知类监控网络的传感器优化部署和数据空间分析类传感网络的传感器稀疏部署；后者主要研究不完全采集的数据空间重构技术和广域监测区域中移动机器人数据采集技术。 本文主要研究内容和贡献如下： 第一，研究了面向事件感知传感器(如监控摄像头)不完全覆盖时的部署位置优化问题。针对监测区域环境复杂，监控对象的信息随空间位置变化频繁的场景。部署在监测区域的监控传感器数目是有限的或者是给定的条件下，为了提高事件的感知能力，本文提出权重感知的传感器不完全覆盖策略，将有限的感知设备集中部署在关键的位置。在这类物联网应用场景中，监测区域各个部分的权重计算是关键，本文将整个监测区域网格化，从感知网格的异常事件发生概率、异常事件影响度、异常事件容忍时间三方面融合计算感知网格的权重。为了方便计算与分析，本文提出拉锯映射方法对网格的权重进行计算。综合传感精度和覆盖率，结合感知网格的权重，提出了监测传感网络的感知可信度的模型，从而保证可靠且有效的感知到监测区域异常事件。 第二，研究了有利于数据重构的传感器稀疏部署问题。针对广域监测区域中传感器感知范围不能完全覆盖监测区域，即监测区域的面积远远大于传感器感知面积，需要考虑传感器稀疏部署的传感器放置问题。本文充分利用监测区域物理量之间的空间相关性，提出了迭代四分格传感器稀疏部署算法。该方法有利于插值重构过程中重构精度的提高。实验表明当利用反距离插值重构数据空间时，本文所提的迭代四分格稀疏部署方法使得部署更加均匀化，具有较高的重构精度。 第三，研究无线传感器网络不完全采集下的鲁棒性数据空间重构策略。针对监测区域中存在部分传感器数据丢失或失真情况下的重构未部署传感器区域的数据空间问题。由于监测的物理量存在空间关联性，本文利用距离非采集点最近的采集点的信息来重构非采集点的信息。传感网络通常部署在条件恶劣的环境中，信息丢失或者失真是不可避免的。本文利用神经网络的鲁棒性，提出学习算子的概念，将反距离插值算法融入BP人工神经网络，并利用反正切函数调和多个预测值，得到具有鲁棒性的数据重构方法。仿真实验表明本文所提算法比反距离插值算法具有较高的重构精度和较强的鲁棒性。 第四，研究了能量和时间受限的移动采集问题。移动机器人采集数据时，机器人能量和执行任务的时间会受限。如何在机器人能量和时间受限条件下最大化采集信息量面临着技术挑战，本文提出能量和时间感知的移动采集策略，将能量和时间感知的移动采集转化成为一个多目标优化问题。采用极大极小法将多目标中的各个目标量进行标准化，从而把多目标融合成单目标问题进行优化求解。本文采用主流的粒子群优化算法进行采集路径的在线规划优化，改进粒子群优化算法，提出了融合认知的粒子群优化算法。为了获得较好的解空间，本文进一步研究了粒子的层层竞争机制，使得粒子群优化算法的搜索空间扩大，收敛较快。通过仿真实验，可以看出本文所提算法比现在主流的粒子群优化算法能获得较早的收敛和较宽的优化解空间，使得机器人在能量和时间受限条件下收集信息量大、遇障平均概率较小、行驶总路程较短。 关键词：物联网；传感器；稀疏部署；移动采集；粒子群优化算法

 Access to information in the physical world is the basis of various applications and services provided by the Internet of Things (IoTs). Most of current research on sensor sensing is mostly based on the assumption of ideal sensing environment or sufficient number of sensors. In fact, the deployment of sensors and data acquisition is often constrained in many application scenarios of the Internet of Things. Due to various factors such as the sensor's sensing ability, deployment cost, and physical environment, it is difficult for the Internet of Things to fully cover the monitoring region. Research on improving the perception of the Internet of Things in the case of constrained sensor deployment and data acquisition can expand the application range of the Internet of Things and become a hotspot in the research of IoTs perception in recent years. In order to improve the sensing capability of the Internet of Things under the constraints of sensor deployment and data acquisition, some key issues in sensor deployment optimization and acquisition mechanism are studied comprehensively and systematically in this paper. The main content is divided into two parts: sensor deployment strategy and data acquisition key technologies. The former mainly studies sensor deployment of event-aware monitoring network and sensor sparse deployment of data-space analysis sensor network; the latter mainly studies data acquisition technology of mobile robot in wide area monitoring area and data space reconstruction technology of entire monitoring regions. These two parts are progressive and interrelated. The main contents and innovations of this paper are as follows: Firstly, the deployment optimization of event-aware sensors (such as surveillance cameras) with incomplete coverage is studied. This section focuses on the case that the change of physical quantity with spatial position is frequent in the complex environment of monitoring region. When the number of sensors deployed in the monitoring region is limited or given, in order to improve the capture rate of events, we adopt a weight-aware sparse deployment strategy. We centrally deploy limited sensing devices in key areas. In such Internet of Things application scenarios, the key is to calculate the weight of each place in the monitoring region. This article grids the entire monitoring region. The weights of sensing grids are calculated from three aspects: the probability of occurrence of anomalous events, the impact of anomalous events and the tolerance time of anomalous events. In order to facilitate calculation and analysis, this paper proposes a method of seesaw mapping to calculate the weights of the grids. Then sensors are deployed in the grid center according to the weights. Combining the sensing accuracy and coverage of sensors, and considering the weights of the sensing grids, a model of sensing reliability of the wireless sensor networks is proposed to ensure reliable and efficient abnormal event perception. Secondly, the research is about the problem of sensor sparse deployment for data reconstruction. When the sensors cannot cover the monitoring region completely, that is, the area of the monitoring region is much larger than the sensing area of all sensors, this paper proposed a sparse sensor deployment algorithm based on iterative dividing sub-regions, taking full advantage of the spatial correlation between the physical quantities of the monitoring region. This method is beneficial to improve the reconstruction accuracy in the process of interpolation reconstruction. Experiments show that the proposed sparse deployment method based on iterative sub-regions can improve the accuracy of reconstruction of the data space when using inverse distance interpolation. The proposed iterative quartile sparse deployment method makes the deployment more uniform and has higher reconstruction accuracy. Thirdly, the robust strategy of data space reconstruction of the locations where the sensors are not deployed under incomplete data acquisition of wireless sensor networks is studied in this paper. This paper focuses on the data reconstruction problem in the case of some sensor data loss or distortion in the monitoring region. Because of the spatial correlation of the physical quantities monitored, this article uses the information of sensor-deployed location closest to the location where the sensor is not deployed to reconstruct its information. Wireless Sensor networks are usually deployed in harsh environments, and information loss or errors are inevitable. In this paper, we use the robustness of neural networks and propose the concept of learning operator. We introduce inverse distance interpolation algorithm into artificial neural network. Using arctangent function to harmonize multiple predictions, a robust data reconstruction method is obtained. The simulation results show that our algorithm has higher reconstruction accuracy than the inverse distance interpolation algorithm. Our algorithm has strong robustness. Fourthly, this paper studies the energy and time-aware mobile acquisition problem. When collecting data for mobile robots, the energy of the robot is limited, and the time for the robot to perform tasks is also limited. How to maximize the amount of information collected under the constraints of energy and time for robots is facing technical challenges. This paper proposed an energy and time-aware mobile acquisition strategy. We transform energy and time-aware mobile acquisition into a multi-objective optimization problem. We use the maximum and minimum method to normalize each target quantity in the multi-objective, so as to integrate the multi-objective into a single-objective problem for optimal solution. In this paper, the popular particle swarm optimization (PSO) algorithm is used to re-plan and optimize the acquisition path. We improved the particle swarm optimization algorithm and proposed a mixed cognitive particle swarm optimization algorithm. In order to obtain optimized results, we also proposed a layer-by-layer screening method for the particle survival of the fittest. In this way, the search space of PSO is enlarged and the convergence rate is faster. Through simulation experiments, we can see that our improved algorithm can achieve earlier convergence and wider optimization solution space than some popular particle swarm optimization algorithms. It makes the robot collect more information, encounter less obstacle probability and travel shorter total distance under energy and time constraints. Key words: internet of things; sensors; sparse deployment; mobile acquisition; particle swarm optimization