Ciuffreda Ilaria, Casaccia Sara, Revel Gian Marco
Department of Industrial Engineering and Mathematical Sciences, Polytechnic University of Marche, 60131 Ancona, Italy.
Sensors (Basel). 2023 Aug 5;23(15):6963. doi: 10.3390/s23156963.
This work illustrates an innovative localisation sensor network that uses multiple PIR and ultrasonic sensors installed on a mobile social robot to localise occupants in indoor environments. The system presented aims to measure movement direction and distance to reconstruct the movement of a person in an indoor environment by using sensor activation strategies and data processing techniques. The data collected are then analysed using both a supervised (Decision Tree) and an unsupervised (K-Means) machine learning algorithm to extract the direction and distance of occupant movement from the measurement system, respectively. Tests in a controlled environment have been conducted to assess the accuracy of the methodology when multiple PIR and ultrasonic sensor systems are used. In addition, a qualitative evaluation of the system's ability to reconstruct the movement of the occupant has been performed. The system proposed can reconstruct the direction of an occupant with an accuracy of 70.7% and uncertainty in distance measurement of 6.7%.
这项工作展示了一种创新的定位传感器网络，该网络使用安装在移动社交机器人上的多个被动红外（PIR）传感器和超声波传感器来在室内环境中定位居住者。所提出的系统旨在通过使用传感器激活策略和数据处理技术来测量运动方向和距离，以重建室内环境中人员的运动。然后，使用监督式（决策树）和无监督式（K均值）机器学习算法对收集到的数据进行分析，以分别从测量系统中提取居住者运动的方向和距离。已经在受控环境中进行了测试，以评估使用多个PIR和超声波传感器系统时该方法的准确性。此外，还对该系统重建居住者运动的能力进行了定性评估。所提出的系统能够以70.7%的准确率重建居住者的方向，距离测量的不确定度为6.7%。
