TopoBrick: Agentic Topology Sampling of Exogenous Variables for Zero-Shot Building IoT Forecasting

2026-07-07 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Internet of Things (IoT) & Connected Devices · Depth: Expert, quick

Summary

TopoBrick is a novel training-free framework designed for zero-shot building IoT forecasting, addressing the limitation of existing forecasters that often treat building sensors as isolated time series or rely on fixed covariate sets. This framework leverages building knowledge graphs to create a compact structural skeleton and employs an agentic topology sampler to intelligently select target-specific exogenous variables. These variables are then organized based on their deployment-time availability, distinguishing between past-known sensor states and future-known calendar, schedule, and meteorological data. Evaluated across three real-world buildings, TopoBrick demonstrates superior performance compared to robust zero-shot foundation-model baselines and achieves competitive results against fully trained building-specific models. Ablation studies confirm that its topology-aware sampling method is significantly more reliable than random, ontology-only, or fixed-hop selection, particularly for physically coupled HVAC and weather-driven sensing variables.

Key takeaway

For Machine Learning Engineers developing building IoT forecasting solutions, consider integrating topology-aware sampling methods like TopoBrick's. Your models can achieve zero-shot performance competitive with fully trained systems by leveraging building knowledge graphs and intelligently selecting exogenous variables based on their availability. This approach significantly enhances reliability, especially for critical HVAC and weather-driven sensor data, reducing the need for extensive training data.

Key insights

TopoBrick uses agentic topology sampling of knowledge graphs for zero-shot building IoT forecasting, outperforming baselines.

Principles

• Building sensors are embedded in physical topology.
• Topology-aware sampling improves IoT forecasting reliability.
• Exogenous variables should be organized by availability.

Method

TopoBrick constructs a compact structural skeleton from building knowledge graphs. It then uses an agentic topology sampler to select target-specific exogenous variables, organizing them by deployment-time availability for zero-shot forecasting.

In practice

• Apply knowledge graphs for sensor context.
• Prioritize topology-aware variable selection.
• Separate past-known from future-known covariates.

Topics

• Building IoT
• Zero-Shot Forecasting
• Knowledge Graphs
• Topology Sampling
• Exogenous Variables
• HVAC Systems

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, AI Engineer

Related on AIssential

• Self-Evolving Multi-Agent Network for Industrial IoT Predictive Maintenance — Hierarchical multi-agent systems with gradient-based policy optimization enable adaptive, efficient, and trustworthy industrial IoT predictive maintenance.
• Token Arena: A Continuous Benchmark Unifying Energy and Cognition in AI Inference — AI inference endpoint performance varies significantly across configurations, impacting cost, energy, and quality.
• TopoEvo: A Topology-Aware Self-Evolving Multi-Agent Framework for Root Cause Analysis in Microservices — TopoEvo uses topology-aware multi-agent reasoning to accurately identify root causes in dynamic microservice environments.
• Zero Touch Predictive Orchestration: Automating Time-Series Models for the Cloud-Edge Continuum — Automated data-mixing with a foundational dataset mitigates cold start for time-series forecasting in volatile Cloud-Edge Continuum environments.
• Vibe Graphing: 10x More Affordable than Vibe Coding (MAS-Factory) — Vibe Graphing automates multi-agent system design from natural language, drastically cutting costs and development complexity.
• Local AI: How & Where To Start Building Something You Can Monetize — Local AI monetizes idle hardware by building workflow-based agents augmented with robust, structured knowledge and memory.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.