Autonomous Room Acoustic Measurements using Rapidly-Exploring Random Trees and Gaussian Processes
Georg Götz, Ishwarya Ananthabhotla, Sebastià V. Amengual Garí, Paul T. Calamia · 2022
Various robot systems have been proposed in the past to automate the tedious and time-consuming room acoustic measurement process.While small-scale measurements within a limited area can be realized with robotic arms, room-scale measurements require robots that can travel larger distances and ideally navigate through their environment autonomously.In this paper, we propose a new measurement strategy for large-scale, autonomous, roomacoustic measurement robots.The measurement strategy uses rapidly-exploring random trees to determine multiple candidate paths, from which it chooses the best path for exploring the unvisited parts of the environment and reconstructing a target acoustic metric.Gaussian process regression is used to incrementally merge new acoustic data into a global estimate.We evaluate the measurement strategy in a multi-room scenario, utilizing a late reverberation metric and a robot system consisting of a source and a receiver robot.We demonstrate that the measurement strategy can be used to map and reconstruct late reverberation characteristics over a large area.