基于旋转超声信号的室内环境建模研究

发布时间：2018-02-28 19:10

本文关键词： 超声环境建模特征提取数据关联模糊理论　出处：《哈尔滨工业大学》2017年硕士论文　论文类型：学位论文

【摘要】：近年来随着服务机器人的快速发展,智能化已成为必然趋势,自主导航作为机器人智能化的重要指标,要求机器人拥有识别周围环境并定位的能力。但是实际生活中机器人所面对的环境大部分无法预先知晓,所以自主识别周围环境并进行地图绘制成为必然需求。基于超声信号的环境建模技术,因其传感器发展最早、最为成熟,小范围测量精度高,对人体没有伤害,价格便宜,转化产品成本低廉等优点,在室内服务机器人上有广阔的应用前景。因此,本文根据室内环境特点,设计数据稳定性强,可以提取多种特征的旋转超声平台,进行环境特征地图构建的研究。在环境特征提取方面,需要面对温度变化、传感器间串扰、超声传感器定向性差、测量误差等影响。本文通过温度补偿抵消不同环境温度导致的测量值偏差。使用4个超声传感器,以正方形结构搭建旋转平台,完成对环境的360°扫描,在保证效率的同时获取更多的环境特征信息,并抑制传感器间的串扰影响。为了优化数据结构剔除误差点,本文提出了基于距离差值的孤立点检测滤波算法,另外针对超声传感器定向性差的问题,提出了相似距离值超声弧聚类算法与以中心线超声模型为基础的Minimum Distance of Point(MDP)特征提取算法,将每个超声弧特征点位置提取出来,并通过MDP几何弧线模型进行特征类型的区分。针对分体式超声传感器环境分辨能力差的问题,提出了模糊MDP特征提取算法,实现了分体式传感器旋转测距情况下的环境特征提取。在地图构建方面,针对收发一体超声传感器,实现了利用MDP算法与最小二乘拟合构建环境地图的仿真实验。针对收发分体超声传感器,提出了切线法,解决了高偏差、极少特征点情况下的直线特征拟合问题,并拥有较高的精度。而对于较大环境的直线拟合,提出了分段切线法,解决了大环境下因估计直线角度偏差造成的误差累计问题。通过模糊MDP算法与切线法拟合完成了环境地图构建,实验结果有较高的精确性。
[Abstract]:In recent years, with the rapid development of service robot, intelligence has become an inevitable trend, autonomous navigation as an important index of robot intelligence, Robots are required to have the ability to identify and locate their surroundings. But in real life, most of the environments that robots face are not known in advance. Therefore, it is necessary to identify the surrounding environment and map it independently. The environmental modeling technology based on ultrasonic signal has the earliest development, the most mature sensor, the high precision of small range measurement, no harm to human body, and low price. The conversion product has the advantages of low cost, so it has a broad application prospect on indoor service robot. Therefore, according to the characteristics of indoor environment, this paper designs a rotating ultrasonic platform with strong data stability, which can extract many features. In the aspect of environmental feature extraction, we need to face temperature change, crosstalk between sensors, poor orientation of ultrasonic sensor. In this paper, the error of measurement is offset by temperature compensation. Four ultrasonic sensors are used to build a rotating platform with square structure, and 360 掳scanning of the environment is completed. In order to optimize the data structure and eliminate error points, an outlier detection filtering algorithm based on distance difference is proposed in this paper. In addition, aiming at the problem of poor orientation of ultrasonic sensors, a similar distance value ultrasonic arc clustering algorithm and a Minimum Distance of point MDP feature extraction algorithm based on the centerline ultrasonic model are proposed to extract the location of each ultrasonic arc feature point. The MDP geometric arc model is used to distinguish the feature types. A fuzzy MDP feature extraction algorithm is proposed to solve the problem of poor environmental resolution of the split ultrasonic sensor. In the aspect of map construction, aiming at receiving and receiving integrated ultrasonic sensor, we realized the environmental feature extraction under the circumstance of split sensor rotating ranging. The simulation experiment of constructing environmental map by using MDP algorithm and least square fitting is realized. Aiming at the transceiver ultrasonic sensor, the tangent method is proposed, which solves the problem of linear feature fitting under the condition of high deviation and few feature points. And it has high precision. For the straight line fitting in larger environment, a piecewise tangent method is proposed. The problem of accumulative error caused by estimating the angle deviation of straight line in large environment is solved. The construction of environmental map is completed by using fuzzy MDP algorithm and tangent method. The experimental results have high accuracy.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242;TP212

【参考文献】