Physically Intelligent Insect-Inspired Antennae for Tactile Perception

Supervisor: Professor Jean-Michel Mongeau

Physically Intelligent Insect-Inspired Antennae for Tactile Perception

This project investigates how insect-inspired physical intelligence can improve robotic tactile sensing. By mimicking the mechanical gradients and active sensing strategies of cockroach antennae, I demonstrated how morphology and dynamics can simplify tactile classification and reduce computational demands in robotics.

🔹 Research Focus

• Explored how stiffness gradients and contact speeds in antenna mechanics shape tactile signal processing.
• Developed bioinspired computational models in MuJoCo and robophysical experiments based on cockroach antennae.
• Generated tactile tensors (spatiotemporal representations of tactile stimuli) under varied contact conditions.
• Validated findings through sim-to-real transfer on a miniature distributed robotic antenna.

Robophysical antenna and its digital twin in MuJoCo

🔹 Key Innovations

• Developed computational models in MuJoCo, serving as a digital twin of both a biological antenna and a distributed tactile sensor system, enabling studies of biomechanics and guiding robotic sensor design.
• Demonstrated that physical intelligence in biological antenna increases tactile data sparsity and dispersion, improving feature classification.
• Identified mechanical gradients and slower contact speeds as critical for enhancing tactile perception.
• Developed a novel tactile tensor framework to represent and classify tactile scenes.

Tactile tensor generation

🔹 Skills & Tools

• Bioinspired design & robotics: translating insect antenna mechanics into robotic prototypes.
• Computational modeling: develop computational models in a physical engine (MuJoCo).
• Machine learning for classification: analyzing spatiotemporal tactile signals for feature recognition.

Software Used:

• Python – MuJoCo, data analysis
• MATLAB – machine learning, signal processing, and experimental data visualization