Contact-force distribution optimization and control for quadruped robots using both gradient and adaptive neural...

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Quadruped robots require sophisticated control for stability, especially when facing external forces.
• Existing methods struggle with dynamic disturbances and uncertainties in complex environments.

Purpose of the Study:

• To develop an optimal feet force distribution and control strategy for quadruped robots under external disturbance forces.
• To enhance robot stability and adaptability in unpredictable conditions.

Main Methods:

• Formulation of constrained dynamics and derivation of a reduced-order dynamical model.
• Utilization of a gradient neural network for optimizing tip-point force distribution under constraints.
• Proposal of a hybrid motion/force control using an adaptive neural network to handle perturbations.

Main Results:

• The proposed method effectively distributes forces and moments on supporting legs to equilibrate external wrenches.
• The adaptive neural network control successfully compensates for environmental perturbations and approximates feedforward forces.
• Simulations verified the control's ability to confront uncertainties, including approximation errors and external disturbances.

Conclusions:

• The developed hybrid motion/force control strategy enhances quadruped robot robustness against external disturbances.
• The integration of gradient and adaptive neural networks offers a promising approach for advanced robotic control.
• This research contributes to more stable and reliable operation of legged robots in dynamic environments.