| The obstacle detection technology based on ultrasonic sensors has been studied for Autonomous Land Vehicle (ALV) applications in this thesis. The thesis consists of three closely related subjects: ultrasonic range-finding system, ultrasonic ranging processing method, map building and obstacle recognition.The validity of correlation detection arithmetic is proved by theory and experiment based on the multicenter ultrasonic signal sampling platform. The pseudorandom code instead of single impulse is used for modulation to improve the capability of anti-jamming and precision of measuring range. Besides, m sequence with single envelope is replaced by m sequence with polar envelopes, which are utilized to realize PSK demodulation for the improvement of the processing gain and capability of anti-jamming.In order to realize the map building based on ultrasonic, a transmitting signal model is established and the D-S theory is used for updating the environment map. The improved ultrasonic model and adaptive range confidence factor are introduced into the range-finding system for decreasing the disturbance caused by specular reflection.The BP neural network is used for classifying three kinds of common obstacle. A model of ultrasonic echo from the surface of obstacle is built, therefore the ultrasonic echo amplitude is simulated with the disturbing of random disturbance. After the training by BP neural network, the range data measured from the moving ultrasonic range-finding system can be used for obstacle recognition. |
PDF Full Text Request