A Multi-Agent System for Building-Age Cohort Mapping to Support Urban Energy Planning

Kundan Thota; Thorsten Schlachter; Veit Hagenmeyer

cs.CVMar 18, 2026

A Multi-Agent System for Building-Age Cohort Mapping to Support Urban Energy Planning

Authors: Kundan Thota, Thorsten Schlachter, Veit Hagenmeyer

Organizations: Institute for Automation and Applied Informatics (IAI), Karlsruhe Institute of Technology (KIT), Karlsruhe, Germany

Abstract

Determining the age distribution of the urban building stock is crucial for sustainable municipal heat planning and upgrade prioritization. However, existing approaches often rely on datasets gathered via sensors or remote sensing techniques, leaving inconsistencies and gaps in data. We present a multi-agent LLM system comprising three key agents, the Zensus agent, the OSM agent, and the Monument agent, that fuse data from heterogeneous sources. A data orchestrator and harmonizer geocodes and deduplicates building imprints. Using this fused ground truth, we introduce BuildingAgeCNN, a satellite-only classifier based on a ConvNeXt backbone augmented with a Feature Pyramid Network (FPN), CoordConv spatial channels, and Squeeze-and-Excitation (SE) blocks. Under spatial cross validation, BuildingAgeCNN attains an overall accuracy of 90.69% but a modest macro-F1 of 67.25%, reflecting strong class imbalance and persistent confusions between adjacent historical cohorts. To mitigate risk for planning applications, the address-to prediction pipeline includes calibrated confidence estimates and flags low-confidence cases for manual review. This multi-agent LLM system not only assists in gathering structured data but also helps energy demand planners optimize district-heating networks and target low-carbon sustainable energy systems.

Figures & tables

Fig. 1: Outline: (1) Multiple data sources (databases, web pages, maps, and PDFs) are ingested into (2) LLM-driven agents to produce (3) geocoded age cohorts.

Fig. 2: Overview of the proposed pipeline for building-age cohort mapping: A multi-agent system for building-age cohort mapping consists of three data‐collection agents (Zensus, OSM, Monument), then fed into a Data Orchestration, Fusion, and Harmonization that integrates and standardizes records into five age cohorts. The BuildingAgeCNN model (ConvNeXt + FPN + CoordConv + SE) is then trained on satellite imagery for supervised learning and inference.

Fig. 3: Inference pipeline of the AgeCohort agent: address parsing, geocoding, satellite-tile retrieval, BuildingAgeCNN inference, and confidence-based flagging for manual review.

Data Source	Coverage Statistics
	Building Count	Coverage (%)
OSM total buildings	72,446	100.00
Fused dataset	15,336	21.17

TABLE I: Coverage of age-labeled buildings in Aachen (sources and fused output).

Cohort	Dataset Statistics
	Count	Share (%)
pre–1919	2,722	17.76
1919–1950	1,114	7.26
1951–1978	10,212	66.60
1979–2000	892	5.82
post–2000	396	2.58

TABLE II: Age-cohort distribution in the fused dataset.

Fig. 4: Representative roof types across age cohorts (left to right: pre-1919, 1919–1950, 1951–1978, 1979–2000, post-2000).

Fig. 5: Data combined into six spatially contiguous folds, with each fold visualized as a region polygon.

Model	Variant	Metrics
		Accuracy (%)	Macro-F1 (%)	Params (M)
ConvNeXt [ 23 ]	Baseline	$82.27\pm 5.55$	$63.71\pm 5.57$	27.82
	+ FPN	$89.06\pm 5.43$	$64.36\pm 5.35$	30.48
	+ FPN + CoordConv	$89.27\pm 3.83$	$66.18\pm 4.72$	31.08
	+ FPN + CoordConv + SE	$\mathbf{90.69}\pm\mathbf{4.22}$	$\mathbf{67.25}\pm\mathbf{5.52}$	31.09
ResNet-50 [ 26 ]	Baseline	$76.07\pm 3.98$	$56.93\pm 3.81$	25.60

TABLE III: Ablation results on the building-age benchmark.

Cohort	Metrics
	Accuracy (%)	F1 (%)
pre–1919	81.96	88.84
1919–1950	24.22	35.71
1951–1978	96.38	89.95
1979–2000	80.64	86.19
post–2000	31.25	35.31

TABLE IV: Per-class performance: accuracy and F1 by age cohort.

Fig. 6: Row-normalized confusion matrix for five cohorts.

Confidence Threshold	Flagged Predictions
	Count	Share (%)
0.65	128	18.28
0.70	172	24.57
0.75	259	37.00

TABLE V: Flagged predictions at different confidence thresholds.

Cohort	Thermal Transmittance (U-values)
	Roof	Upper Ceiling	Wall	Floor
pre–1919	1.95	1.00	2.10	2.05
1919–1950	1.54	1.06	1.41	1.41
1951–1978	0.75	0.96	1.05	1.29
1979–2000	0.42	0.39	0.52	0.62
post–2000	0.24	0.25	0.28	0.33

TABLE VI: Cohort-averaged U-values (W/m 2 K) for envelope components from TABULA reference data.

Remote Sensing Image Understanding

Remote sensing image understanding uses AI to interpret satellite and aerial imagery for objects, land cover, change, hazards, and spatial relationships.

See all 208 papers

Building age estimation based on widely available multi-source1 geospatial data
Oct 12, 2026Jialu Sui, Hongsheng Zhang, Rui Liu +4
Department of Geography, The University of Hong Kong, Hong Kong, China · Institute for Climate and Carbon Neutrality, the University of Hong Kong, Hong Kong, China · The University of Hong Kong Shenzhen Institute of Research and Innovation, Shenzhen, China +5
SENSE: Satellite-based ENergy Synthesis for Sustainable Environment
May 18, 2026Kailai Sun, Mingyi He, Heye Huang +5Conditional Image Generation Synthetic Data Generation
SMART, Singapore · Massachusetts Institute of Technology Cambridge, MA, USA · University of Florida Gainesville, Florida, USA
Multi-Agent Collaborative Reasoning with Tool-Augmented Evidence for Urban Region Profiling
Jul 15, 2026Xixuan Hao, Yutian Jiang, Jiabo Liu +4Multi-Agent Reasoning Multimodal Agents
The Hong Kong University of Science and Technology (Guangzhou) Guangzhou, China · University of Washington Seattle, USA · Chang’an University Xi’an, China +1

Oct 12, 2026J/K move · Enter open · S save

Building age estimation based on widely available multi-source1 geospatial data

Jialu Sui, Hongsheng Zhang, Rui Liu, Jing Ling+3

Department of Geography, The University of Hong Kong, Hong Kong, China · Institute for Climate and Carbon Neutrality, the University of Hong Kong, Hong Kong, China · The University of Hong Kong Shenzhen Institute of Research and Innovation, Shenzhen, China+5

Open paper AskPDF arXiv

Building age is fundamental for urban infrastructure safety, operation, and planning, as it directly affects structural safety, and sustainable development. Existing methods mostly rely on detailed building information such as street view, function, or materials, which are harder to obtain than geospatial data like satellite imagery and Digital Surface Models (DSMs). However, existing methods based on those geospatial data mainly rely on optical imagery and local texture features, limiting their ability to capture structural information and distinguish buildings from different periods. Moreover, the inherent temporal ordering among building age categories is often overlooked. To address these limitations, we propose the first deep-learning framework for multi-building age estimation using widely available multi-source geospatial data, including remote sensing data and their derived products, as well as building footprint information that estimates age at both pixel and instance levels within a unified pipeline. Specifically, a large pretrained backbone with fine-tuning adapters extracts representative features, and raster-derived and property-derived fusion modules support pixel-wise and instance-wise predictions. Pixel-level estimation employs a pyramid decoder with Cosine-guided Mixture-of-Experts (CMoE) to fuse multi-source raster data across scales, capturing fine-grained building age differences, while instance-level prediction uses attention pooling and Property-Conditional Adaptive Instance Normalization (PCAdaIN) to extract instance-level representations and integrate property attributes. A Temporal Ordinal Constraint Loss (TOCL) enforces the sequential ordering of predicted ages. We also construct a building age benchmark for Hong Kong to enable comprehensive evaluation. The source code and datasets are available at https://github.com/littlebeen/RS-buildingAge.git.

Abstract

Remote Sensing Image Understanding

Explore similar work

Building age estimation based on widely available multi-source1 geospatial data

SENSE: Satellite-based ENergy Synthesis for Sustainable Environment

Multi-Agent Collaborative Reasoning with Tool-Augmented Evidence for Urban Region Profiling