Biped Robot Ankle Torque Control for Underactuated Walking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Biped robots face challenges in controlling their motion, particularly in complex and rugged terrains, due to underactuation of the support legs, which limits control over the center of mass and floating base during dynamic movements.
Innovation Solution
A method and apparatus for controlling biped robots by acquiring the support phase, center of mass speeds, and torque vectors of ankle joints, using quadratic programming (QP) for whole-body motion control, ensuring complete control over the robot's motion and adapting to both fully actuated and underactuated states.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional control methods are used for biped robots, then the control system is simpler, but the control over center of mass and floating base is insufficient during dynamic movements
Solution Approach 1:
The control method segments the biped robot control into distinct phases (support phase and swing phase) and calculates torque vectors for different joints (ankle, knee, hip) separately. This segmentation allows the complex control problem to be broken down into manageable parts while maintaining complete control over the center of mass and floating base during dynamic movements.
Solution Approach 2:
The control method dynamically adjusts the torque vectors based on the robot's current state, including position, velocity, and acceleration of the center of mass. The quadratic programming model continuously optimizes the torque distribution according to the dynamic conditions, enabling adaptive control that maintains reliability while managing system complexity.
2Adaptability or versatility
If quadratic programming is used for whole-body motion control, then complete control over motion is achieved, but the computational complexity increases
Solution Approach 1:
The control method pre-calculates the relationship between torque vectors and center of mass motion through dynamic models. By preparing the quadratic programming formulation in advance with pre-computed mass matrices and Coriolis terms, the method enables rapid online optimization while maintaining complete adaptability for whole-body motion control.
Solution Approach 2:
The quadratic programming approach changes the control parameters dynamically by adjusting the torque vectors based on the desired center of mass trajectory and current robot state. This parameter adaptation allows the system to achieve versatile motion control while the structured formulation manages computational complexity through efficient optimization.
3Adaptability or versatility
If the robot adapts to both fully actuated and underactuated states, then the robot can operate in more conditions, but the control difficulty increases
Solution Approach 1:
The control method uses a universal quadratic programming framework that can handle both fully actuated and underactuated states through the same mathematical formulation. By expressing the control problem in terms of torque vectors and constraint forces that are valid for both cases, the method achieves multi-functionality without increasing control difficulty.
Solution Approach 2:
The control method incorporates feedback from the robot's current state (position, velocity, acceleration) to continuously adjust the torque vectors. This feedback mechanism enables the system to automatically adapt to different operational conditions, whether fully actuated or underactuated, while maintaining ease of operation through closed-loop control.
Data Source
AI summary
A method for controlling a biped robot includes: acquiring a support phase of the biped robot in a current state; acquiring a current forward speed and a current lateral speed of a center of mass of the biped robot at an end of the support phase; acquiring a torque vector of an ankle joint of a support leg of the biped robot according to the current forward speed and the current lateral speed; acquiring a target torque vector according to the torque vector and a preset Quadratic Programming (QP) model; and controlling the biped robot to move according to the target torque vector.