Quadrupedal Robot Locomotion: A Biologically Inspired Approach and Its Hardware Implementation.

Area of Science:

• Robotics
• Computational Neuroscience
• Bio-inspired Systems

Background:

• Robotic locomotion systems often require complex control mechanisms.
• Spiking Neural Networks (SNNs) offer a biologically plausible approach to generating complex motor patterns.
• Central Pattern Generators (CPGs) are neural circuits responsible for rhythmic motor activities.

Purpose of the Study:

• To develop and validate a bio-inspired locomotion system for quadruped and hexapod robots.
• To utilize Spiking Neural Networks (SNNs) as Central Pattern Generators (CPGs) for robot locomotion.
• To employ a metaheuristic method for optimizing SNN parameters for effective locomotion.

Main Methods:

• A bio-inspired locomotion system was designed using Spiking Neural Networks (SNNs) as Central Pattern Generators (CPGs).
• SNN parameters (synaptic weights and topologies) were optimized using Christiansen Grammar Evolution (CGE), a metaheuristic method.
• The system was implemented and validated on physical quadruped and hexapod robot platforms, including simulations of leg amputation.

Main Results:

• The SNN-based CPG system successfully generated diverse locomotion patterns for both quadruped and hexapod robots.
• The system demonstrated effective control on a quadruped robot using an Arduino board (35% resource usage).
• The system showed high efficiency on a hexapod robot using a Spartan 6 FPGA (3% resource usage), even with simulated leg loss.

Conclusions:

• The proposed bio-inspired locomotion system effectively controls robotic platforms.
• Spiking Neural Networks (SNNs) provide an efficient and adaptable control mechanism for complex robotic locomotion.
• The metaheuristic optimization approach is effective in configuring SNNs for robust robotic locomotion across different platforms and conditions.