Semantic Context Matching for Transferable Building Automation Models

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing building automation systems face challenges in adapting to changes in context, as pre-trained evaluation and control models are not easily transferable to new setups with different sensor configurations or data formats, leading to inefficiencies in deployment and adaptation.

Innovation Solution

The method employs semantic matchmaking by generating semantic descriptions of both the target and source contexts, allowing for the selection and adaptation of pre-trained models through a confidence-based matching process, enabling the transfer and adaptation of models to new contexts with recommended modifications.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Loss of time

If pre-trained evaluation and control models are used for building automation systems, then the initial setup time is reduced, but the models cannot be easily transferred to new contexts with different sensor configurations or data formats

Engineering Contradiction:
Improvemodel training timeVSAvoidmodel transferability
Core Design Contradiction:
Loss of timeVSAdaptability or versatility

Solution Approach 1:

The patent applies preliminary action by generating semantic descriptions of the context in which pre-trained models were trained before deployment. This includes creating detailed representations of sensor configurations, data formats, and environmental conditions. When a model needs to be transferred to a new context, the semantic description of the target context is generated and compared with source context descriptions, allowing the system to pre-assess compatibility and prepare appropriate adaptations before actual deployment, thus reducing both training time and transfer difficulties.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent utilizes parameter changes by representing context information in a standardized semantic format that can be systematically compared and adapted. The semantic descriptions include adjustable parameters such as sensor types, data formats, and environmental conditions. When transferring models between contexts, the system identifies parameter differences and applies transformations to align the source model's semantic description with the target context, enabling flexible model adaptation without complete retraining.

Inventive Principle:
Principle #35Parameter changes

2Adaptability or versatility

If semantic descriptions are generated for context matching, then model transferability between different contexts is improved, but the complexity of the system increases

Engineering Contradiction:
Improvemodel transferabilityVSAvoidsystem complexity
Core Design Contradiction:
Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent applies universality by creating a standardized semantic description framework that serves multiple functions: it characterizes training contexts, describes target contexts, enables compatibility assessment, and guides model adaptation. This universal semantic representation approach allows the same mechanism to handle diverse context types (different sensors, data formats, environmental conditions) without requiring separate specialized systems for each context type, thus improving transferability while controlling complexity through standardization.

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

3Reliability

If manual configuration of building automation devices is performed, then system reliability is improved, but the setup time becomes extremely time-consuming

Engineering Contradiction:
Improvesystem configuration reliabilityVSAvoidsetup time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

The patent applies feedback by using semantic context descriptions to automatically evaluate whether pre-trained models are suitable for target contexts. The system compares source and target context semantics, provides feedback on compatibility, and automatically adjusts or selects appropriate models. This feedback mechanism enables the system to maintain reliable configurations by validating semantic compatibility while dramatically reducing setup time compared to manual configuration, as the automated feedback loop quickly assesses and adapts models without requiring time-consuming manual verification.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP3851922A1A method for providing an evaluation and control model for controlling target building automation devices of a target building automation system
Publication Date: 2021.07.21 ABB (SCHWEIZ) AG
  • EP3851922A1 patent drawingFigure 1~2
  • EP3851922A1 patent drawingFigure 3
  • EP3851922A1 patent drawing

AI summary

A method for providing an evaluation and control model for controlling target building automation devices of a target building automation system, wherein a plurality of pre-trained source evaluation and control models for controlling source building automation devices of a source building automation system is provided, and wherein for each pre-trained source evaluation and control model a semantic based description of the context in which the model was trained is generated, is characterized in that a semantic based description of the context of the target building automation system is generated, a semantic based description of the context of each pre-trained source evaluation and control model of said plurality of pre-trained source evaluation and control models is retrieved, and the generated semantic based description of the context of the target building automation system and the semantic based description of the context in which the pre-trained source evaluation and control models were trained are matched by using a semantic matchmaking concept.