多传感器数据融合及其在吸尘机器人避障中的应用

发布时间：2018-07-24 19:47

【摘要】：自动检测及复杂系统自动控制总免不了多传感器数据采集与信息的综合运用。因此,作为一种重要的信息处理方法,异质传感器的数据融合成为当前的一个研究热点。本文分别使用传统数学算法和计算智能算法对超声和红外传感器的数据融合问题进行了研究。在编程仿真的基础上,构建了一个模糊神经网络,研究了多传感器数据融合在吸尘机器人自主避障中的应用。本文主要做了以下几方面的工作:(1)使用传统数学算法对异质传感器数据融合进行了研究。本文提出了一种自适应加权融合算法,对超声和红外传感器测距系统的距离信息进行融合。使用MATLAB对提出的算法进行了编程仿真,仿真结果表明:本文提出的自适应加权融合算法融合结果稳定,算法收敛速度快,能根据传感器的方差大小对各传感器的权值进行自适应分配。(2)使用计算智能算法、构建BP神经网络对异质传感器的数据融合进行了研究。本文针对超声和红外传感器数据融合,构造了一个三层BP神经网络,使用MATLAB对设计的BP神经网络进行了编程仿真。为解决仿真结果中出现的标准BP神经网络收敛速度慢,融合结果不够理想的问题,提出了增加附加动量项的训练算法,并对其改进效果进行了对比分析,结果显示:较之于普通网络,附加动量项的BP神经网络收敛速度更快,融合结果更加精确。(3)对数据融合技术在吸尘机器人避障中的应用进行研究。针对具有五个超声测距传感器和一个角度传感器的扫地机器人原型系统,构建了一个五层模糊神经网络控制系统。将五个超声传感器测得的距离和一个角度传感器测得的目标方位角作为系统的输入变量,输入模糊神经网络系统,经过系统的推理计算得到网络输出。使用MATLAB对设计的模糊神经网络系统进行了仿真,仿真结果表明:当环境无障碍物时,扫地机器人可以从设置的起点沿直线运动到目标点;当环境中存在障碍物时,扫地机器人可以有效避开障碍物到达目标点。
[Abstract]:Automatic detection and automatic control of complex systems can not avoid multi-sensor data acquisition and comprehensive application of information. Therefore, as an important information processing method, data fusion of heterogeneous sensors has become a hot topic. In this paper, traditional mathematical algorithms and computational intelligence algorithms are used to study the data fusion of ultrasonic and infrared sensors. On the basis of programming and simulation, a fuzzy neural network is constructed, and the application of multi-sensor data fusion in autonomous obstacle avoidance of dust cleaning robot is studied. The main work of this paper is as follows: (1) the data fusion of heterogeneous sensors is studied by using traditional mathematical algorithms. In this paper, an adaptive weighted fusion algorithm is proposed to fuse the range information of ultrasonic and infrared sensor ranging systems. The proposed algorithm is programmed and simulated with MATLAB. The simulation results show that the fusion result of the proposed adaptive weighted fusion algorithm is stable and the convergence speed of the algorithm is fast. The weights of each sensor can be allocated adaptively according to the variance of the sensor. (2) A BP neural network is constructed to study the data fusion of heterogeneous sensors. In this paper, a three-layer BP neural network is constructed for the data fusion of ultrasonic and infrared sensors. The designed BP neural network is programmed and simulated by MATLAB. In order to solve the problem of slow convergence speed and unsatisfactory fusion result of standard BP neural network in the simulation results, a training algorithm to increase the additional momentum term is proposed, and the improved results are compared and analyzed. The results show that the convergence speed of BP neural network with momentum term is faster and the fusion result is more accurate than that of ordinary network. (3) the application of data fusion technology in obstacle avoidance of dust cleaning robot is studied. A five-layer fuzzy neural network control system for a floor sweeping robot with five ultrasonic ranging sensors and an angle sensor is constructed. The distance measured by five ultrasonic sensors and the target azimuth angle measured by one angle sensor are taken as the input variables of the system and the fuzzy neural network system is inputted and the network output is obtained by the inference calculation of the system. The design of fuzzy neural network system is simulated by MATLAB. The simulation results show that when there are no obstacles in the environment, the floor sweeping robot can move along a straight line from the setting point to the target point, and when there are obstacles in the environment, The floor sweeping robot can effectively avoid obstacles to reach the target point.
【学位授予单位】：西安理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP242

【参考文献】