Acoustic Echo Sensing for Robot Localization in Buried Pipe Networks

Abstract

Robot localization in buried pipes presents many challenges, including the unavailability of sensing methods, such as a global positioning system, and the limited perspective of sensors, such as vision. This article addresses these challenges by using acoustic sensing, where sound emitted by robots propagates long distances and around corners in the pipe environment and is used to estimate the distance to reflective features. The reverberant environment produces many echoes that contain useful information for localization, but it also produces many second-order echoes that have reflected from two features consecutively, which creates a challenge for localization. Therefore, a novel approach for using first- and second-order echoes in pose-graph optimization for localization is developed here. Experiments demonstrated that the acoustic echo localization approach gave a relative estimate error of 0.24% (with 0.23% standard deviation). We compared to the results of a Kalman filter localization algorithm (3.4% error) and a pose-graph optimization alternative designed for only first-order echoes (0.80% error) and also compared to results from sensing approaches in the literature including vision sensing (2.17%–2.27% error), acoustic sensing with a separate transmitter (3.5% error), and radio beacon sensing (0.047% error). This suggests that acoustic echo sensing surpasses these alternative methods, except those that require external hardware, and gives a tenfold improvement on other approaches without external hardware. In summary, we have developed a new method for echo localization in pipes, incorporating second-order echoes into pose-graph optimization, and demonstrated its effectiveness with respect to alternative approaches.