Statistical estimation algorithms for ultrasonic detection of surface features

For several applications of interest in the robotic field, range sensing is often accomplished by means of single ultrasonic sensors. Additional information about the objects´ surface, such as the orientation relative to the sensing device, has to be extracted by means of ultrasonic sensor arrays. So far, a few researchers have dealt with the problem of determining also the local curvature of curved reflectors. The aim of this paper is to investigate the minimum amount of range information needed to estimate distance, orientation and radius of cylindrical targets by means of a linear array of ultrasonic transducers. The proposed imaging problem is stated within the field of statistical estimation theory: an iterative linearised least-squares estimator, i.e. an extended Kalman filter, is designed and used for optimal processing of the sensed data. The theoretical analysis of the filtering algorithm performance allows one to elucidate the critical factors ultimately affecting the achievable accuracies of the proposed technique. Experimental results are presented, in close agreement with the predictions of the theoretical analysis