Optogenetically enhanced physical reservoir computing with in vitro neural networks for obstacle avoidance

Area of Science:

• Neuroscience
• Robotics
• Artificial Intelligence

Background:

• Reservoir computing utilizes recurrent neural networks for complex tasks.
• Optogenetics allows precise control of neuronal activity.
• Neuro-robotic systems integrate biological components with robotic control.

Purpose of the Study:

• To investigate the impact of optogenetic stimulation on in vitro neural network performance for obstacle avoidance.
• To develop a neuro-robotic system using optogenetically controlled neural networks and the FORCE learning algorithm.

Main Methods:

• An all-optical biological reservoir computing framework was employed, integrating optogenetics and calcium imaging.
• A closed-loop system was developed, utilizing the FORCE learning algorithm to guide a virtual agent.
• Neuronal activity was precisely regulated and recorded to control the system's behavior.

Main Results:

• The system achieved high accuracy and efficiency in obstacle navigation, with over 95% success rate.
• Optogenetic stimulation significantly improved the obstacle avoidance success rate and system adaptability.
• Stable task control was achieved with a minimal output from just 15 neurons.

Conclusions:

• Optogenetically controlled biological neural networks show great potential for neuro-robotic applications.
• Physical reservoir computing with optogenetics enables accurate and efficient obstacle avoidance.
• This framework could advance brain-inspired intelligence and neuro-robotic control systems.