Optimization strategies to obtain smooth gait transitions through biologically plausible central pattern generators

Area of Science:

• Computational Neuroscience
• Robotics
• Biophysics

Background:

• Central pattern generators (CPGs) are neural networks crucial for generating animal locomotion.
• Modeling CPGs for gait generation and transition requires biologically accurate, high-dimensional neuron and synapse models.
• Parameter tuning in these complex nonlinear systems is challenging due to unpredictable emergent dynamics.

Purpose of the Study:

• To explore global optimization strategies for parameter optimization in multigait CPG models.
• To develop and optimize parameters for a novel, minimal-topology, interlimb hexapod CPG model.
• To investigate a biologically plausible mechanism for gait transitions compatible with short-term synaptic plasticity.

Main Methods:

• Utilized global optimization strategies for parameter tuning of CPG models.
• Employed an existing quadruped CPG model as a test bed for objective function formulation.
• Developed and optimized parameters for a new hexapod CPG model with complex cell dynamics.

Main Results:

• Successfully optimized parameters for a hexapod CPG model, achieving functional gaits.
• Demonstrated prompt gait transitions in the hexapod model by solely adjusting control currents.
• Maintained fixed CPG parameters post-optimization, indicating robust parameter tuning.

Conclusions:

• Global optimization is effective for tuning complex CPG models with minimal topology.
• Control current modulation offers a viable mechanism for rapid gait transitions in CPGs.
• The optimized CPG model and transition mechanism align with principles of short-term synaptic plasticity.