Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster A in Poster Session A: Tuesday, August 4, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Emergent Functional Circuits in Embodied Foragers

Salar Nouri¹, Harikrishna Satheesh Pillai¹, Pakhi Banchalia², Naoto Yoshida³, Roxana Zeraati⁴, Shervin Safavi¹; ¹Technische Universität Dresden, ²Amrita Vishwa Vidyapeetham (Deemed University), ³Kyoto University, ⁴Max Planck Institute for Biological Cybernetics

Presenter: Salar Nouri

Effective brain-body interactions are crucial for survival, as the brain monitors internal bodily states to regulate homeostasis and balance energy with cognitive and motor functions. However, the neural mechanisms linking physiological states to neural dynamics and behavior remain poorly understood. To explore this connection, we developed an embodied artificial neural network (ANN) agent to adaptively forage in a realistic environment, while maintaining physiological homeostasis. The agent was trained using homeostatic reinforcement learning, which applies a homeostatic objective instead of simply maximizing reward. We found that the agent learns to forage adaptively in the environment, i.e., it adapts its exploration and movement velocity to its satiety state, with broader and faster exploration occurring at higher hunger levels, consistent with recent observations in rodents. Using behaviorally guided analysis of neural dynamics, we identified distinct clusters of neurons encoding movement, internal state, and action inhibition. Our findings offer a novel computational framework to uncover the role of brain-body interactions in adaptive behaviors.

Topic Area: Methods, Tools, Theory & Neural Coding