Local AI Occupant Control for Privacy-Safe Building Automation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing building management systems lack an economical and ecological solution for controlling actors in living or working units that is tailored to individual occupant needs while protecting personal data, due to legal restrictions on data sharing and the burden of manual configuration or complex automation.
Innovation Solution
A decentralized AI system, such as the 'Vibe' apparatus, processes personal occupant data locally using removable storage elements with security layers, enabling personalized control of building actors without internet connectivity, and employs federated learning to adapt to individual habits while keeping data private.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If centralized cloud-based automation systems are implemented, then living comfort and appliance efficiency are improved, but data privacy protection and system complexity increase
Solution Approach 1:
The system segments the automation architecture into distributed edge devices deployed in individual living units, each processing data locally, rather than a centralized cloud system. This segmentation enables personalized automation while keeping data private to each unit, resolving the contradiction between comfort improvement and data privacy protection.
Solution Approach 2:
The patent introduces federated learning as an intermediary mechanism that allows edge devices to learn from each other without sharing raw personal data. The system uses encrypted model updates and aggregation protocols as mediators, enabling collective intelligence while maintaining data privacy, thus improving comfort without compromising privacy.
2Loss of information
If manual control of appliances is used, then data privacy is maintained, but energy efficiency and living comfort deteriorate
Solution Approach 1:
The system enables appliances to self-regulate based on locally processed data from sensors and occupant behavior patterns. Edge devices automatically optimize energy consumption without manual intervention or external data sharing, allowing energy efficiency improvement while maintaining data privacy through local autonomous decision-making.
Solution Approach 2:
The patent implements local feedback loops where edge devices continuously monitor appliance performance and occupant behavior, then automatically adjust settings to optimize energy efficiency. This closed-loop control system improves energy efficiency through automated adaptation while keeping all data processing local, preserving data privacy.
3Ease of operation
If personalized automation is implemented, then living comfort is improved, but system complexity and installation burden increase
Solution Approach 1:
The edge devices are designed to self-configure and automatically adapt to each living unit's specific needs through local learning. The system eliminates complex manual installation and configuration by enabling devices to autonomously personalize automation, thereby improving comfort while reducing the perceived complexity for end users.
Solution Approach 2:
The patent implements dynamic adaptation where the automation system continuously learns and adjusts to changing occupant behaviors and preferences. This dynamic personalization occurs automatically through local machine learning, providing tailored comfort without requiring complex static configuration or frequent manual reconfiguration.
4Loss of energy
If landlords invest in efficient automation solutions, then energy efficiency is improved, but cost increases and data access restrictions prevent optimization
Solution Approach 1:
The system segments the automation investment into modular edge devices deployed at individual living unit level rather than a expensive centralized system. This segmentation reduces per-unit cost while enabling personalized energy optimization, allowing landlords to invest efficiently without requiring access to tenants' personal data.
Solution Approach 2:
The patent uses federated learning as an intermediary that enables cost-effective optimization without centralized data collection. The system aggregates learning from multiple edge devices through encrypted model updates, allowing landlords to benefit from collective intelligence and reduced energy consumption across the portfolio without incurring high data infrastructure costs or violating privacy restrictions.
Data Source
Figure 1~2a
Figure 2b
Figure 3
AI summary
The invention concerns a system comprising - a first apparatus comprising -- access means for accessing at least one storage element for storing personal occupant data, wherein the access means comprise a receptacle configured to removably hold the storage element and are protected by at least one security layer, -- means for locally processing of at least parts of the personal occupant data by adjusting a configuration of at least one local machine learning model using at least the personal occupant data -- means for controlling at least one actor of the living or working unit based on the local processing and - at least one remote apparatus comprising -- means for communicating with at least two local apparatuses, configured to obtain at least a respective part of a respective at least one locally adjusted configuration of a respective local machine learning model from the at least two local apparatus -- means for remote training, configured to determining a remotely adjusted configuration of at least one of the local machine learning model using the at least two locally adjusted configurations, and -- wherein the means for communicating are further configured to provide at least a part of the remotely adjusted configuration to at least one of the at least two local apparatus.