Approaching Sound Object with Sensorimotor Coordination when Sensors Partially Damaged

Sound source localization is becoming a fundamental ability for robots, which makes use of audio clues. However, few studies have considered how to leverage the motion of the robot to locate the sound source more precisely, especially when the robot’s acoustic sensors are partially impaired. In this paper, we propose a new method for the mobile robot to better approach the sound object based on the idea of sensorimotor coordination. Firstly, the optimal direction near fine microphone sensors will be found to defeat the influence of some bad acoustic sensors. Secondly, the direction of the sound object is estimated with the HOA encoding technique using a spherical microphone array. Thirdly, the robot actively rotates to refine the position of the sound source combining the estimated direction and odometry information in a coordinated way considering sensors partially damaged. And the robot is approaching the target while the position of the sound object is calculated iteratively. Experimental results in the simulated and real-world environment show the robot can approach the sound object more accurately using the proposed method.