Robot Controller Configuration Using Simulation-Trained Agents

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing methods for configuring robot controllers for predefined tasks are inefficient and require manual programming, which is time-consuming and lacks flexibility.

Innovation Solution

A method involving machine learning-based training of an agent using robot and environmental parameters in simulations, followed by configuring the robot's control system based on the trained agent, utilizing reinforcement learning and neural networks, with optional user input and cloud-based simulations to enhance efficiency and accuracy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If manual programming is used to configure robot controllers, then the configuration can be performed with simple tools, but the process is time-consuming and inefficient

Engineering Contradiction:
Improveconfiguration speedVSAvoidprogramming time
Core Design Contradiction:
ProductivityVSLoss of time

Solution Approach 1:

The patent replaces manual programming operations with machine learning-based automatic generation. The system uses trained ML models to automatically generate robot controller configurations from task descriptions and environmental models, substituting the mechanical process of manual coding with an automated intelligent system that significantly reduces configuration time and increases productivity

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

Solution Approach 2:

The system enables self-service configuration where the robot controller automatically generates its own control programs based on input parameters and trained machine learning models. The configuration process becomes self-directed, requiring minimal human intervention and eliminating the need for manual programming expertise, thereby dramatically reducing the time loss associated with traditional configuration methods

Inventive Principle:
Principle #25Self-service

2Adaptability or versatility

If traditional configuration methods are used, then the system structure remains simple, but the system lacks flexibility and adaptability

Engineering Contradiction:
Improveconfiguration flexibilityVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies preliminary action by pre-training machine learning models with extensive robot parameters, environmental models, and task scenarios before actual deployment. This pre-computation phase creates adaptable configuration systems that can quickly respond to various tasks without requiring complex real-time processing, thereby increasing flexibility while managing system complexity through advance preparation

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The system leverages parameter changes by utilizing diverse robot parameters (kinematic, dynamic, geometric) and environmental model parameters as inputs to the machine learning models. The trained models learn to generate appropriate controller configurations by processing these varying parameters, enabling flexible adaptation to different robot types, environments, and tasks without increasing inherent system complexity

Inventive Principle:
Principle #35Parameter changes

3Manufacturing precision

If machine learning training is performed with comprehensive robot and environmental parameters, then task execution accuracy improves, but the training process becomes more complex and resource-intensive

Engineering Contradiction:
Improvetask execution accuracyVSAvoidtraining process complexity
Core Design Contradiction:
Manufacturing precisionVSDevice complexity

Solution Approach 1:

The patent applies segmentation by dividing the comprehensive training process into distinct modules: robot parameter processing, environmental model processing, simulation environment creation, and model training. This modular approach allows each component to be developed and optimized independently, managing training process complexity while maintaining the ability to process comprehensive parameters for high task execution accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system introduces simulation environments as an intermediary between parameter input and model training. The simulation environment serves as a mediator that processes comprehensive robot and environmental parameters in a controlled virtual setting before deploying to real robots, thereby improving task execution accuracy while managing training complexity through layered processing

Inventive Principle:
Principle #24Intermediary (Mediator)

Data Source

PatentEP3760390B1Performance of a predetermined task using at least one robot
Publication Date: 2026.02.11 KUKA DEUT GMBH
  • EP3760390B1 patent drawingFigure 1~3

AI summary

A method according to the invention for configuring a controller (2) of a robot (1) to perform a predetermined task comprises the steps: - Acquiring (S10, S20) at least one robot parameter and at least one environmental model parameter; - Training (S40) an agent using at least one simulation based on the acquired robot parameter and environmental model parameter by means of machine learning based on a predetermined cost function; and - Configuring (S50) the controller of the robot based on the trained agent.