Least Entropy-Like Approach for Reconstructing L-Shaped Surfaces Using a Rotating Array of Ultrasonic Sensors
This work addresses a domain-specific problem in robotics or sensing for accurately reconstructing orthogonal surfaces, presenting an incremental improvement over existing methods.
The paper tackles the problem of reconstructing L-shaped surfaces from ultrasonic sensor data by introducing a new algorithm that classifies data using waveform energy and minimizes a nonlinear cost function inspired by Gibbs entropy to estimate plane parameters robustly against outliers from multiple reflections. Experimental results show improved precision and reliability compared to the classic Least Squares Method.
This paper introduces a new algorithm for accurately reconstructing two smooth orthogonal surfaces by processing ultrasonic data. The proposed technique is based on a preliminary analysis of a waveform energy indicator in order to classify the data as belonging to one of the two flat surfaces. The following minimization of a nonlinear cost function, inspired by the mathematical definition of Gibbs entropy, allows to estimate the plane parameters robustly with respect to the presence of outlying data. These outliers are mainly due to the effect of multiple reflections arising in the surfaces intersection region. The scanning system consists of four inexpensive ultrasonic sensors rotated by means of a precision servo digital motor in order to obtain distance measurements for each orientation. Experimental results are presented and compared with the classic Least Squares Method demonstrating the potentiality of the proposed approach in terms of precision and reliability.