Quantifying the biomimicry gap in biohybrid robot-fish pairs

Area of Science:

• Robotics
• Ethology
• Artificial Intelligence

Background:

• Biohybrid systems integrating robotic components with live animals offer novel approaches to study collective animal behavior.
• A significant challenge is the 'biomimicry gap,' where discrepancies between simulations and reality lead to unrealistic animal responses.

Purpose of the Study:

• To develop and validate a biomimetic robotic lure and a neural network (NN) model for generating realistic social interactions in fish.
• To investigate the effectiveness of biohybrid systems in bridging the biomimicry gap for studying animal behavior.

Main Methods:

• A biomimetic lure of a rummy-nose tetra fish (Hemigrammus rhodostomus) was created.
• A neural network (NN) model was employed to generate biomimetic social interactions.
• Experiments were conducted using a biohybrid pair (fish and robotic lure), a pair of real fish, and simulations.

Main Results:

• The biohybrid system demonstrated social interactions comparable to those of genuine fish pairs.
• The robotic lure and NN showed minimal deviation in real-world interactions versus simulations and fish-only experiments.
• The NN efficiently controlled the robot in real-time, validating its efficacy.

Conclusions:

• Biohybrid systems can effectively generate realistic social interactions, mirroring natural animal behavior.
• Bridging the biomimicry gap through comprehensive validation is essential for the reliable use of these systems.
• This study validates the use of robotic lures and NN for studying collective animal behavior.