A Transputer-based Neural Network To Determine Object Positions From Tri-aural Ultrasonic Sensor Data

For mobile robot navigation purposes, we use an object position in measurement system. The measurement system is composed of two parts, a signal acquisition part and the signal processing part. The first part consists of three ultrasonic sensors, one transmitting and three receiving. In the second part, the topic of this paper, we use a transputer based neural network to process the acoustical signals from the receivers. We shotu how to combine ultrasonic sensor data with an artificial neural network to estimate object positions (distance rscr; and bearing Theta;). Acoustic signals from the three sensors are preprocessed ana! then passed through the neural network. The object positions are determined by the neural network. Experimental results on object position measurement show that a neural network of the multilayer perceptron type can be trained to do this task, and that the resulting system is reasonably sensitive and accurate.