Biped Robot Ankle Torque Control for Unknown Surface Angles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional ankle control algorithms for biped robots fail to account for errors due to differences between the sole of the foot and the support surface, and between intended and actual control inputs, leading to instability and imbalance when stepping on unknown angles.
Innovation Solution
A method that compensates the target torque with an acting torque measured by a sensor and feeds back the actual torque applied by the ankle, using error compensation coefficients Kc and Kd derived from Linear Quadratic Regulator (LQR) control, to calculate the input torque, thereby ensuring balance and conformity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional target torque control is used, then the control system is simple, but the robot cannot maintain balance when stepping on unknown angles
Solution Approach 1:
The patent implements feedback control by measuring the actual torque applied by the ankle joint using a sensor and feeding it back to the control system. The control input torque is calculated as u = τc*Kd - τ*Kc, where τc is the acting torque from the support surface, τ is the actual torque applied by the ankle, and Kd and Kc are error compensation coefficients. This feedback mechanism allows the robot to adjust to unknown surface angles and maintain balance dynamically.
Solution Approach 2:
The patent introduces error compensation coefficients Kc and Kd that are derived from Linear Quadratic Regulator (LQR) control theory. These parameters dynamically adjust the control torque based on the difference between desired and actual states, allowing the system to adapt to varying conditions while maintaining a relatively simple control structure.
2Measurement precision
If error compensation is implemented, then the balance control accuracy is improved, but the control calculation complexity increases
Solution Approach 1:
The patent uses LQR-derived coefficients Kc and Kd to compensate for errors in torque measurement and control. These coefficients are calculated once based on system dynamics matrices (A, B, C) and weight matrices (Q, R), then applied continuously in the control law u = τc*Kd - τ*Kc. This approach provides high precision error compensation without requiring complex real-time calculations.
Solution Approach 2:
The error compensation coefficients are pre-calculated using LQR control theory before the robot operates on unknown surfaces. This preliminary computation of optimal gain matrices allows the robot to handle various error conditions efficiently during operation without performing complex real-time optimization, thus improving precision while keeping runtime complexity manageable.
Data Source
AI summary
Disclosed herein is a system and method of controlling one or more ankles of a walking robot. In the above system and method, an input torque u to be applied to an ankle of a robot is obtained by compensating a target torque τd with an acting torque τc measured using a sensor and applied by a support surface and then an actual torque τ applied by the ankle of a foot is instead fed back into the control system. Therefore, proper balance of the walking robot is ensured.