基于神经网络的移动机器人多传感器数据融合研究

摘要

针对单超声波的测距缺陷，采用多个超声波结合红外开关共同测距，提高整体测量精度；针对BP神经网络训练收敛速度慢，容易陷入局部极小值等缺点，加入动量-自适应因子来改进BP神经网络；将改进的BP应用于移动机器人传感器旅行家II号数据融合中，实践证明，经改进后的BP神经网络收敛精度高误差小，融合后的信息比未经融合的信息更精确。%To solve the problem of the drawbacks of single ultrasonic distance measurement, a common ranging method of the multiple ultrasonic combined with the infrared switch was presented, and the accuracy of measurement was improved by using this method; Regarding the shortcomings of BP neural network (BPNN) training convergence speed and easily falling into local minima, the momentum adaptive factor was used to improve the BPNN, and then the improved BPNN was applied to the UP-VoyagerⅡmobile robot in data fusion . Practical application shows that,the fused data is more accurate than the unfused data.