Acoustic homing with Bayesian ring attractor network.

In this talk, I will present my current research implementing a Bayesian ring attractor model, originally proposed to explain heading representation in the fly brain, on a physical robot for real-time acoustic homing. The project uses a microphone array, IMU, and camera to estimate the direction of a sound source, feeding these noisy, time-varying signals into a ring attractor network whose activity bump encodes both a heading estimate and a live measure of confidence in that estimate. Using a ROS2 pipeline to coordinate the robot with a Vicon motion-capture system, I evaluate this approach against ground truth to assess the model's performance in a highly noisy environment. The work highlights the practical challenges of translating a neuroscience-inspired probabilistic model into real-time robotic control.

Louis Renard is a Bachelor's student in Computer Science at McGill University in Montreal, Canada. He is currently completing a summer research internship in the Collective Robotics Group at the University of Konstanz, supervised by Dr. Andreagiovanni Reina. At McGill, he has worked as a teaching assistant for Computer Systems and Discrete Structures courses. His broader interests include algorithm design, NLP, and neuroscience-inspired approaches to engineering.