Sniffing, Physiology, and Search: Odor Navigation in Embodied Neural Networks

Area of Science:

• Computational neuroscience
• Robotics
• Animal behavior

Background:

• Olfactory navigation is crucial for many animals.
• Respiration rhythmically gates sensory input, creating complex odor landscapes.
• Understanding how agents process temporally fluctuating signals is key.

Purpose of the Study:

• To investigate respiratory odor navigation in a minimal brain-body-environment model.
• To explore how evolved agents locate chemical sources using breathing-modulated olfactory signals.
• To identify the neural and behavioral mechanisms underlying this navigation.

Main Methods:

• Evolved continuous-time recurrent neural network agents in a simulated environment.
• Utilized a minimal brain-body-environment model.
• Analyzed agent behavior and internal dynamics using dynamical systems analysis.

Main Results:

• Evolution reliably produced agents capable of locating and remaining near the chemical source.
• Identified two robust navigational regimes: continuous oscillatory casting and episodic saltatory search.
• Demonstrated synchronization between respiration and motor output in successful agents, mimicking rodent navigation.

Conclusions:

• A respiratory-neural mechanism supports flexible and efficient odor-guided navigation.
• The agent's respiratory pattern generator dynamically reconfigures motor output for different search strategies.
• Findings provide insights into the principles of sniff-synchronized navigation and sensory-motor integration.