Walking Robot Torque Control for Natural Human-Like Gait
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current torque-based and position-based control methods for walking robots face challenges in achieving natural gait and balance, with torque-based Finite State Machine (FSM) control methods requiring complex dynamic equations for balancing and position-based Zero Moment Point (ZMP) control methods resulting in high energy inefficiency and unnatural gait.
Innovation Solution
A control method that generates reference walking trajectories based on human walking data, using a walking change factor to adjust stride and velocity, and calculates control torques to track target walking trajectories, allowing for various walking patterns and natural gait similar to humans.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If position-based ZMP control method is used, then precise position control is achieved, but high energy consumption and high joint stiffness occur
Solution Approach 1:
The patent replaces the position-based control mechanism with a torque-based control mechanism. Instead of controlling joint positions through high-gain servo systems that consume significant energy, the invention directly controls joint torques to achieve the desired walking motion. This substitution of control paradigm reduces energy consumption while maintaining control precision, as torque control allows for softer, more energy-efficient actuation that naturally follows the dynamics of human-like walking.
2Measurement precision
If position-based ZMP control method is used, then precise position control is achieved, but unnatural gait with bent knees occurs
Solution Approach 1:
The patent replaces position-based control with torque-based control to enable natural human-like gaits. By controlling torques rather than positions, the system allows joints to move more freely and naturally, producing stretched-knee walking patterns similar to humans. The torque control approach incorporates dynamic models that capture the natural dynamics of human walking, enabling the robot to generate biomechanically plausible motions without the artificial constraints of position tracking.
3Use of energy by moving object
If torque-based FSM control method is used, then low energy consumption and natural gait are achieved, but complex dynamic equations are required for balancing
Solution Approach 1:
The patent applies preliminary action by pre-computing and storing optimal torque profiles for various walking conditions and operating states. Instead of solving complex dynamic equations in real-time during walking, the system prepares torque commands in advance based on the current state and desired transitions. This offline computation approach reduces online computational complexity while maintaining the benefits of torque-based control for energy efficiency and natural gait generation.
Solution Approach 2:
The patent employs dynamic state transitions and adaptive torque adjustment to simplify control. By modeling the walking system as a dynamic system with discrete states and transitions, the controller can switch between predefined torque profiles based on current state, avoiding the need to continuously solve complex differential equations. This dynamic approach maintains natural walking patterns while reducing computational burden through state-based decision making.
4Ease of operation
If torque-based FSM control method is used, then natural gait is achieved, but robot balance control is insufficient
Solution Approach 1:
The patent implements feedback mechanisms that monitor the robot's actual state during walking and adjust torque commands accordingly. By continuously comparing the desired state transitions with actual system response, the controller can compensate for deviations and maintain balance. This feedback loop ensures that while the robot maintains natural human-like gaits through torque-based control, it also achieves stable balance by adapting to real-time conditions and correcting deviations from the planned walking pattern.
Data Source
AI summary
A walking robot and a control method thereof. The control method includes storing angle change data according to time corresponding to at least one joint unit of the robot using human walking data, extracting reference knot points from the angle change data according to time, and generating a reference walking trajectory using the extracted reference knot points, calculating a walking change factor to perform change between walking patterns of the robot, generating a target walking trajectory through an arithmetic operation between the reference walking trajectory and the calculated walking change factor, calculating a control torque to track the generated target walking trajectory, and transmitting the calculated control torque to the at least one joint unit so as to control walking of the robot, thereby achieving various walking patterns through a comparatively simple arithmetic operation process.