Enhancing Legged Robot Locomotion Through Smooth Transitions Using Spiking Central Pattern Generators

Area of Science:

• Robotics
• Computational Neuroscience
• Control Systems

Background:

• Legged robots require adaptable locomotion for diverse environments.
• Smooth transitions between gaits are crucial for stability and energy efficiency.
• Spiking neural networks (SNNs) offer bio-inspired control for complex behaviors.

Purpose of the Study:

• To develop a mechanism for smooth, imperceptible gait transitions in a hexapod robot.
• To maintain robot balance and improve energy efficiency during locomotion pattern changes.
• To enable dynamic, terrain-adaptive gait selection.

Main Methods:

• Utilized a spiking neural network (SNN) as a Central Pattern Generator (CPG) for walk, jog, and run gaits.
• Employed SPIKE-synchronization metric to determine optimal gait transition points.
• Integrated FSR sensors to detect terrain rigidity for dynamic gait adjustments.
• Implemented real-time testing on a physical hexapod robot across four terrain types.

Main Results:

• Successfully demonstrated smooth and stable transitions between walking, jogging, and running gaits.
• Achieved near-imperceptible gait changes, enhancing overall locomotion fluidity.
• Showcased improved energy efficiency by minimizing abrupt actuator movements.
• Validated the system's adaptability to different terrains through real-time experiments.

Conclusions:

• The proposed SNN-based CPG with SPIKE-synchronization enables robust and efficient gait transitions in hexapod robots.
• The system's ability to adapt to terrain rigidity enhances its practical applicability.
• The methodology is extensible to other legged robotic platforms for advanced locomotion control.