The New Elman ANN Application in Accuracy Improvement of Robot Navigation and Obstacle Avoidance Technology

Abstract

This paper, through the analysis of the limitation of the ultrasonic sensor in the underground drilling robot which obstacle avoidance, puts forward the solving scheme, emphasizes the compensation of temperature and humidity to ultrasonic sensor, attempts to use Elman feedback neural network approach function. Elman network hidden layer adopts Tan-sigmoid transfer function, and output layer transfer function adopts linear function, which guarantee that the network can approaches wanton function with wanton precision once it has enough layers. Proved by experiments that the measure precision of ultrasonic ranging raises two orders of grade after carrying on the temperature and humidity compensate, improves the working efficiency of obstacle avoidance in the system greatly and the ability of drilling robot’s obstacle avoidance.