基于神经元网络的超声传感器补偿算法及在井下机器人避障中的应用  

Ultrasonic Sensor Compensation Algorithm and Its Application in Pit Robot Obstacle Avoidance System Based on Neural Network

作　　者：付华[1] 杜晓坤[1] 陈峰[1] 

机构地区：[1]辽宁工程技术大学电气工程系,辽宁阜新123000

出　　处：《传感技术学报》2006年第2期511-514,共4页Chinese Journal of Sensors and Actuators

基　　金：辽宁省教育厅基金项目(2004C011)

摘　　要：详细分析了超声波传感器在测距应用方面的局限性,并针对每个局限性提出了解决方案。文章重点指出对超声波传感器进行温度、湿度的补偿,尝试用Elman反馈神经元网络逼近函数。Elman网络隐层采用了“tansig”激活函数,输出层选用了“pureline”激活函数,保证了只要有足够多的隐层神经元个数,网络就可以任意精度逼近任意函数。经实验验证,在对超声波测距传感器进行温度、湿度补偿后,其测量精度提高了两个数量级,大大提高了超声测距传感器的工作精度。在井下机器人避障系统中测试应用后,提高了避障系统的测量精确度,降低了避障系统的误判率,使井下机器人工作安全系数得到了提高。This paper analyzed the limitation of the ultrasonic sensor in the ranging application aspects, put forward the scheme solved according to the limitations. To emphasize that, carry on warm and humidity compensation to ultrasonic wave sensor. Try to use Elman feedback neural network to approach function. Elman network latent layer adopt ＂ tansig ＂ activate function, output layer activate function with ＂ pureline ＂, this guarantee that once the network has enough layers, it can approaches wanton function with wanton precision. Proved by experiment that ultrasonic ranging＇s measure precision raise two orders of magnitude after carried on the warm and humidity compensate. Improved the working efficiency of ultrasonic ranging sensor. After the compensate algorithm is used in the pit robot obstacle avoidance system, the measure accuracy of avoidance system is improved and the safe coefficient is raised, the error rate is decaded as well.

关 键 词：井下机器人 避障 超声测距 ELMAN 

分 类 号：TP242.2[自动化与计算机技术—检测技术与自动化装置]