Autonomous Factory Routing With Local Feedback Between Adjacent Systems

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing solutions, including single-agent Reinforcement Learning (RL) and direct multi-agent Reinforcement Learning (Marl), fail to achieve optimal routing and processing of workpieces in autonomous factories due to limitations in modularity, scalability, and the requirement for global information.

Innovation Solution

A decentralized multi-agent Reinforcement Learning (Marl) approach with Independent Learning (IL) variant is implemented, where each system uses local information to adapt its machine learning model, allowing for modular, scalable, and heterogeneous system compositions without requiring global information.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If decentralized multi-agent reinforcement learning with Independent Learning variant is used, then modularity and scalability are improved, but solution quality deteriorates with increasing number of systems

Engineering Contradiction:
ImprovemodularityVSAvoidsolution quality
Core Design Contradiction:
Adaptability or versatilityVSManufacturing precision

Solution Approach 1:

The autonomous factory is divided into independent plants, each with its own machine learning model that learns locally. This segmentation allows each plant to be trained and deployed independently, improving modularity and scalability while maintaining acceptable solution quality through local optimization

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each plant uses local information from its environment to train its machine learning model, rather than requiring global information. This local quality approach enables decentralized learning where each system adapts to its specific context, improving both modularity and maintaining solution quality through localized optimization

Inventive Principle:
Principle #3Local quality

2Adaptability or versatility

If decentralized multi-agent reinforcement learning with Independent Learning variant is used, then system heterogeneity and scalability are improved, but training efficiency deteriorates due to linear scaling

Engineering Contradiction:
ImprovescalabilityVSAvoidtraining time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

Training is segmented into independent plant-level processes rather than centralized training. Each plant trains its own machine learning model using local data, which allows parallel training execution and reduces the linear time penalty associated with increasing system size

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Each plant independently trains and optimizes its own machine learning model using local environmental information, without requiring centralized coordination or global information. This self-service approach enables autonomous adaptation and reduces training overhead as the system scales

Inventive Principle:
Principle #25Self-service

3Reliability

If direct multi-agent reinforcement learning is used, then coordination between systems is improved, but computational complexity and information requirements worsen

Engineering Contradiction:
ImprovecoordinationVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The complex coordination requirements of direct multi-agent reinforcement learning are extracted and replaced with simpler local learning processes. Each plant independently learns from local feedback signals, eliminating the need for complex inter-agent communication and coordination mechanisms while maintaining system reliability

Inventive Principle:
Principle #2Taking out (Extraction)

Data Source

PatentEP4546065A1Method for operating an autonomous factory having at least two systems that are adjacent to one another, computer program product, computer-readable storage medium, and autonomous factory
Publication Date: 2025.04.30 SIEMENS AG
  • EP4546065A1 patent drawingFigure 1
  • EP4546065A1 patent drawingFigure 2
  • EP4546065A1 patent drawing

AI summary

The invention relates to a method for operating an autonomous factory (10) with at least two adjacent systems (12, 14), comprising the steps of: providing a first system (12) with a first production plan (24), wherein the first system (12) includes a first machine learning model (20); providing a second system (14) with a second production plan (26), wherein the second system (14) includes a second machine learning model (22), and wherein the second system (14) is adjacent to the first system (12); transmitting a workpiece (18) from the first system (12) to the second system (14); generating a feedback message (38) by means of the second system (14); and transmitting the feedback message (38) from the second system (14) to the first system (12) by means of a communication device (48) of the second system (14). The invention further relates to a computer program product, a computer-readable storage medium, and an autonomous factory (10).