Neuro-Mechanical Computational Model of Insect Antennae

Supervisor: Professor Jean-Michel Mongeau

🧠 Neuro-Mechanical Computational Model of Insect Antennae

This project develops a neuro-mechanical model of the cockroach antenna to study how mechanical signals from antennal deflections are transformed into neural activity for tactile feature perception. By coupling physics-based simulations with neural encoding models and electrophysiology experiments, I showed how antenna mechanics shape sparse, spike-based neural codes that enable efficient feature classification.

🔹 Research Focus

• Modeled antenna mechanics as a multi-joint kinematic chain within a physics-based simulation environment (MuJoCo).
• Derived spatiotemporal strain patterns (tactile tensors) from antennal deflections under different contact conditions.
• Developed phenomenological models of campaniform and hair sensilla to encode mechanical strain into spike trains.
• Validated model predictions against extracellular recordings from cockroach antennal nerves.
• Trained spiking neural networks (SNNs) on simulated spike trains to classify tactile features.

Electrophysiology experiment and spike sorting

🔹 Key Innovations

• Created a digital twin of the cockroach antenna linking mechanics to neural encoding.
• Demonstrated strong correlation between model-predicted firing activity and experimental electrophysiology.
• Achieved efficient tactile feature discrimination using sparse, spike-based codes.

🔹 Skills & Tools

• Biomechanics & modeling: physics-based simulation of antenna dynamics (MuJoCo).
• Neural modeling: phenomenological encoding of mechanosensory units.
• Electrophysiology: extracellular nerve recordings, spike sorting, and validation.
• Machine learning: SNN training and tactile feature classification.

Software Used:

• MuJoCo – physics-based antenna simulation
• MATLAB & Arduino & Basler Carema- electrophysiology experiment and control
• OpenEphys & NeuroNexus - ephys dac and CM32 probe
• SpikeInterface & Kilosort4 & phy – spike sorting of neural recordings
• Python – electrophysiology data processing and analysis
• Pytorch & SpikingJelly – SNN training and tactile feature classification