Servo Controller Gain Adjustment via Reinforcement Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing servo control systems fail to adjust controller gains appropriately in response to changes in motor phase, leading to uneven motor rotation due to constant gain settings, which is not easily corrected as the motor's phase changes rapidly.
Innovation Solution
A machine learning device and method that perform reinforcement learning to adjust controller gains based on real-time motor phase calculations, using action and state information to update the action-value function and optimize coefficients for position, speed, and current controllers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If the controller gain remains constant, then the control system is simple to operate, but the motor rotates unevenly due to inductance changes with phase
Solution Approach 1:
The controller gain is transformed from a static constant value to a dynamic variable that changes with motor phase. The gain is calculated using transfer functions where coefficients vary according to the instantaneous motor phase angle, enabling the control system to adapt to inductance changes and maintain uniform motor rotation throughout the rotation cycle
Solution Approach 2:
The controller gain parameters (proportional gain Kp, integral gain Ki, differential gain Kd) are modified as functions of motor phase angle. By changing these parameters dynamically based on phase position, the system compensates for inductance variations and eliminates rotation unevenness while maintaining operational simplicity through automated calculation
2Manufacturing precision
If the controller gain is adjusted according to motor phase, then the motor rotation uniformity is improved, but the device complexity increases due to phase-dependent gain calculation
Solution Approach 1:
The mechanical complexity of phase-dependent gain adjustment is replaced by mathematical computation. Transfer functions with phase-dependent coefficients calculate the optimal gain values automatically based on motor phase angle, eliminating the need for complex mechanical adjustment mechanisms while achieving smooth motor rotation
Solution Approach 2:
The control system performs self-adjustment by automatically calculating phase-dependent gain values through transfer functions. The system monitors its own state (motor phase angle) and autonomously modifies controller parameters without external intervention, reducing the need for complex external control mechanisms
3Manufacturing precision
If the controller gain is adjusted in real-time according to motor phase, then the motor rotation uniformity is improved, but the calculation complexity increases due to instantaneous phase changes
Solution Approach 1:
The transfer functions are pre-configured with phase-dependent coefficient structures before operation. By establishing the mathematical relationships in advance, the system reduces real-time calculation burden to simple parameter substitution based on current phase angle, enabling rapid gain adjustment without complex instantaneous computations
Solution Approach 2:
The controller gain adjustment follows the periodic nature of motor rotation. Gain parameters are updated periodically according to the rotation cycle, with transfer functions calculating appropriate values at different phase positions. This periodic approach simplifies computation by leveraging the repetitive pattern of motor operation
Data Source
AI summary
A machine learning device that performs reinforcement learning with respect to a servo control apparatus that controls target device having a motor, including: outputting action information including adjustment information of coefficients of a transfer function of a controller gain to a controller included in the servo control apparatus; acquiring, from the servo control apparatus, state information including a deviation between an actual operation of the target device and a command input to the controller, a phase of the motor, and the coefficients of the transfer function of the controller gain when the controller operates the target device based on the action information; outputting a value of a reward in the reinforcement learning based on the deviation included in the state information; and updating an action-value function based on the value of the reward, the state information, and the action information.