Abstract
Animals provide unique sources of inspiration for robotics, particularly species that rely on non-visual sensing modalities to navigate complex environments. Among them, bats constitute a remarkable model system: they achieve agile flight and coordinated collective motion by relying exclusively on echolocation even within dense swarms where numerous overlapping acoustic signals give rise to the so-called Cocktail Party Problem. Amazed by the ability of bats to navigate and coordinate collectively through sound alone, we developed RO-BAT, a bio-inspired robotic platform designed to investigate how sound-based sensing can be translated into robotic systems for navigation and swarm coordination. The first stage of the project focused on passive sound localisation, in which the robot localises external sound sources without emitting its own signals. Drawing inspiration from bats’ ability to estimate the direction of arrival (DOA) of acoustic cues, a microphone array was integrated with a Raspberry Pi and a sound card to enable real-time, onboard signal processing for obstacle avoidance. Three DOA algorithms were implemented and evaluated in controlled experiments, achieving reliable obstacle avoidance in dynamic multi-agent scenarios. This work established a proof of concept for sound-only navigation in robots, demonstrating particular value when visual sensing is unreliable or unavailable. Building on this foundation, the second stage implemented active echolocation, equipping the RO-BAT with an ultrasonic transducer for emission and a custom-built microphone array optimised for ultrasound localisation. A signal processing pipeline was developed to extract both range and angular information from echoes. Several DOA algorithms were tested and Delay-and-Sum (DAS) was selected for its computational efficiency. Laboratory experiments confirmed that the robot could autonomously avoid obstacles using only its self-generated acoustic cues, effectively reproducing the core principles of bat echolocation on a robotic platform. The current implementation of a single functioning RO-BAT—developed in both its passive and active configurations—is now being scaled up into a swarm of autonomous, decentralised agents. This swarm of individually controlled RO-BATs is already capable of navigating cluttered environments acoustically, without relying on any other sensory cues. By modelling both passive and active acoustic localisation, the RO-BAT platform demonstrates a biologically inspired approach to robotic navigation and swarm coordination. Its design highlights how bats’ natural strategies for managing overlapping acoustic signals can lead to scalable, resource-efficient sensing solutions for robotics. This work not only contributes to the study of collective bioacoustics in natural systems, but also paves the way for robust swarm robotic navigation in challenging environments.
About the speaker
Alberto Doimo is a Master's student in Acoustic Engineering at Politecnico di Milano, specialised in audio signal processing and echolocation. He is also a contributor to The Sound of Bees project at CASCB, to develop a way to record the Stop signal inside beehives.