Neuromechanical Simulation with NEURON and MuJoCo.

Area of Science:

• Computational Neuroscience
• Robotics
• Biophysics

Background:

• Existing simulation platforms lack integrated tools for modeling brain, body, and environment concurrently.
• Bridging the gap between neural activity and physical embodiment is crucial for understanding complex behaviors.

Purpose of the Study:

• To present a novel method for simulating neuromechanical models.
• To demonstrate the integration of NEURON and MuJoCo for comprehensive brain-body-environment simulations.
• To showcase different neural control strategies for musculoskeletal models.

Main Methods:

• Utilized a combination of NEURON for neural modeling and MuJoCo for physics simulation.
• Developed open-loop models with simple spiking neurons and realistic motoneurons.
• Implemented a closed-loop central pattern generator model with physics engine feedback.

Main Results:

• Successfully simulated neuromechanical models integrating neural control with musculoskeletal dynamics.
• Demonstrated the feasibility of controlling realistic musculoskeletal models using NEURON-generated neural activity.
• Validated the closed-loop model with a central pattern generator and sensory feedback.

Conclusions:

• The integrated NEURON-MuJoCo framework provides a powerful tool for neuromechanical simulations.
• This approach advances computational neuroscience by enabling holistic brain-body-environment modeling.
• The presented models serve as a foundation for future research in embodied artificial intelligence and neuroscience.