Audio-Visual Trajectory Planning in Dynamic Environment and Navigation Assist System for Autonomous System

Abstract

Autonomous systems pose several opportunities and challenges at the same time in many civilian and military applications. Auditory system is an essential ingredient in robotics and Sound Source Localization (SSL) is an important part of such system. Such system may become essential when other sensing systems (vision for example), are impaired due to bad lighting conditions or any other reason. The auditory system is even expected to be economical and small enough to fit on the robot which adds more difficulties to the constraints of accuracy and robustness. In this work, a two-stage approach to the acoustic localization problem is suggested. The aim is to minimize the search area for the SRP-PHAT algorithm and increase the reliability and accuracy of the localization system especially when using low cost compact microphone array. The search area is minimized by estimating the Direction of Arrival (DoA) of the acoustic location and then forming a boundary around this estimated DoA according to the confidence level of this estimation along with the range of the microphone array. The Root Mean Square Error (RMSE) obtained with the proposed approach is lower than SRP-PHAT algorithm in more than 90% of the cases. The results obtained also proved that the proposed approach is successful even when only 1000 search points are used instead of 125000 points in case of the conventional algorithm, which greatly reduces the time of calculations to be less than 1% of time taken with conventional algorithm.