Robot Control Parameter Learning for Servo Gain and Motion Tuning

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current robotic systems face challenges in setting operation parameters, particularly in optimizing performance parameters such as servo gains, acceleration, and deceleration characteristics, which are difficult to adjust accurately through artificial means, leading to suboptimal robot performance and increased complexity in operation.

Innovation Solution

A control device equipped with a calculation unit using machine learning to observe state variables like torque and position information, and a learning portion that optimizes operation parameters based on these observations, evaluating rewards for performance metrics like time, accuracy, vibration, and sound levels to automatically adjust parameters for improved robot performance.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If operation parameters are artificially determined, then the setting process is simple, but the robot performance cannot be sufficiently extracted

Engineering Contradiction:
Improverobot performanceVSAvoidparameter setting complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The robot system performs self-optimization of operation parameters through machine learning. The calculation unit automatically determines optimal parameters by learning from observed state variables (torque, position) and reward signals, enabling the system to improve its own performance without external intervention. This resolves the contradiction by making the system self-sufficient in parameter optimization while maintaining simplicity in deployment.

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system dynamically changes operation parameters (servo gains, acceleration, deceleration) based on learned patterns from state observations. By continuously adjusting these parameters through machine learning optimization, the system extracts maximum robot performance while the parameter changes are automatically managed, reducing the apparent complexity for users.

Inventive Principle:
Principle #35Parameter changes

2Productivity

If machine learning is used to optimize parameters, then robot performance is maximized, but the system complexity increases

Engineering Contradiction:
Improveoperation efficiencyVSAvoidcontrol system complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The calculation unit serves multiple functions: it observes state variables, processes torque and position information, performs machine learning computations, evaluates reward signals, and determines optimal operation parameters. By consolidating these diverse functions into a single multi-functional unit, the system achieves high operation efficiency while managing complexity through functional integration rather than proliferation of separate components.

Inventive Principle:
Principle #6Universality (Multi-functionality)

Solution Approach 2:

The calculation unit acts as an intermediary between the robot's physical state (observed through sensors) and the control system (which executes commands). It translates raw state variables into optimized operation parameters through machine learning, serving as a intelligent mediator that bridges perception and action while encapsulating the complexity of the learning algorithm within this intermediate layer.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If artificial parameter setting is used, then the system is easy to implement, but servo gains and acceleration characteristics cannot be accurately adjusted

Engineering Contradiction:
Improveparameter adjustment accuracyVSAvoidparameter setting ease
Core Design Contradiction:
Measurement precisionVSEase of operation

Solution Approach 1:

The system implements feedback by observing the robot's actual state variables (torque, position) during operation and using this information to evaluate reward signals. The calculation unit continuously refines operation parameters based on this feedback loop, achieving high precision in servo gain and acceleration characteristics adjustment. The feedback mechanism automatically handles the complexity of precise tuning, making the system both accurate and easy to operate.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent replaces manual mechanical adjustment of parameters with an automated computational system. Instead of physically tuning servo gains and acceleration characteristics through manual intervention, the system uses machine learning algorithms to compute optimal values based on observed state variables. This substitution of mechanical/manual adjustment with computational optimization achieves superior precision while maintaining ease of operation.

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

4Ease of operation

If remote parameter adjustment is used, then high-grade know-how is required, but the difficulty level remains high

Engineering Contradiction:
Improveparameter setting easeVSAvoidknowledge requirement
Core Design Contradiction:
Ease of operationVSAdaptability or versatility

Solution Approach 1:

The robot system performs self-optimization of operation parameters through machine learning. The calculation unit automatically determines optimal parameters by learning from observed state variables (torque, position) and reward signals, enabling the system to improve its own performance without external intervention. This resolves the contradiction by making the system self-sufficient in parameter optimization while maintaining simplicity in deployment.

Inventive Principle:
Principle #25Self-service

Data Source

PatentEP3357651B1Control device, robot, and robot system
Publication Date: 2022.07.13 SEIKO EPSON CORP
  • EP3357651B1 patent drawingFigure 1
  • EP3357651B1 patent drawingFigure 2
  • EP3357651B1 patent drawingFigure 3

AI summary

A control device includes a calculation unit that calculates an operation parameter related to an operation of a robot by using machine learning, and a control unit that controls the robot on the basis of the calculated operation parameter.