SPEAKER: Usama Saqib, PostDoc fellow at ITU.
ABSTRACT: Robot audition is a subdomain of robotics that combines robotic platforms with advanced audio signal processing techniques. This field enables a robot to interact with its environment and specifically with humans, paving the way for Human-Robot Interaction (HRI). The domain was popularized by Japanese researchers in the 1990s and early 2000s.
In this talk, we explore a new area of research within robot audition that enables a robot to create a spatial map of its environment primarily using acoustic echoes. Echolocation is a concept that has been adapted by animals, and even some humans, to “sense” their surroundings; it has been studied extensively to create sonar-based systems. However, creating spatial maps using airborne sound presents significant physical challenges. For instance, the speed of sound in air is roughly four times slower than in water (approx. 343 m/s vs. 1,480 m/s), which complicates precise map generation.
In this work, we investigate combining several spatial filtering techniques, such as beamforming, with Time-of-Arrival (TOA) estimators to generate an accurate environment map. Moreover, we explore the use of audible frequencies instead of traditional ultrasonic frequencies, a significantly more difficult problem due to ambient noise and interference. This research could potentially pave the way for developing lower-cost robotic platforms for factory environments using off-the-shelf components.