Radius of curvature and location estimation of cylindrical objects with sonar using a multi-sensor configuration

Abstract

Despite their limitations, sonar sensors are very popular in time-of-flight measuring systems since they are inexpensive and convenient. One of the most important limitations of sonar is its low angular resolution. An adjustable multi-sonar configuration consisting of three transmitter/receiver ultrasonic transducers is used to improve the resolution. The radius of curvature estimation of cylindrical objects is accomplished with this configuration. Two different ways of rotating the transducers are considered. First, the sensors are rotated around their joints. Second, the sensors are rotated around their centers. Also, two methods of tirne-of-flight estimation are implemented which are thresholding and curve-fitting. Sensitivity analysis of the radius of curvature with respect to some important parameters is made. The bias-variance combinations of both estimators are compared to the Cramer-Rao lower bound. Theory and simulations are verified by experimental data from real sonar systems. Data is smoothed by extended Kalman filtering. Rotating around the center works better than rotating around the joint. Curve-fitting method is shown to be better than thresholding method both in the absence and presence of noise. The best results are obtained w'hen the sensors are rotated around their centers and the curve-fitting method is used to estimate the time- of-flight. There is about 30% improvement in the absence of noise and 50% improvement in the presence of noise.