Published October 10, 2026 | Version v1

Code, trained policies and derived results for: Quantifying the indoor air quality benefit of schedule-aware proactive control learned by reinforcement learning

  • 1. ROR icon Korea Institute of Machinery & Materials
  • 2. ROR icon Kyung Hee University

Description

Environment, PPO training, evaluation and analysis code, the 776 trained policies (698 proactive and comparison runs and 78 reactive RL runs retrained for this study), the exported weights of the 30 paired CO2 and PM2.5 PINN surrogates, and the derived results for the study of schedule-aware proactive control of windows, doors and an air purifier in a small timetabled space. The surrogate networks are archived with the air-purifier study (Song et al., Building and Environment 306 (2026) 115390, https://doi.org/10.1016/j.buildenv.2026.115390; data and code at https://doi.org/10.5281/zenodo.22074863). See README.md for the layout. Code is released under the MIT License (LICENSE-CODE); policies, surrogate weights and results under CC BY 4.0.

Files

code_si_evidence_v1.zip

Files (53.7 MB)

Name Size
md5:5f80dd22eb9aa7b4fbbb75cf37bb244d
20.8 kB Preview Download
md5:8bb87b180f75c51a9132c53cd806e9f6
152.4 kB Preview Download
md5:a674c6be7231d63d1e082f3e2c785d20
9.0 MB Preview Download
md5:31da0e62ab90fcb6ae3b12a4e4d5c119
9.0 MB Preview Download
md5:21d59a355eaa6d1247c9d974a649c8f2
9.0 MB Preview Download
md5:1f0ed12475afb4ba83bc29e31ece262c
9.0 MB Preview Download
md5:6da055e7312a4cbe561bc85e01aceabf
9.0 MB Preview Download
md5:a2d3194f89ddabc2f9c8fba0047aefe9
2.8 MB Preview Download
md5:38e0a041f59d2a2bcc5fbb6e30fdef06
5.2 MB Preview Download
md5:b0534a7b0ee8c0fd6522627b311440a0
3.5 kB Preview Download
md5:7cd5d8ba5ca3a26a4a75d046aafbfb6c
472.0 kB Preview Download

Additional details

Related works

References
Dataset: 10.5281/zenodo.22074863 (DOI)
Journal article: 10.1016/j.buildenv.2026.115390 (DOI)
Journal article: 10.1016/j.buildenv.2026.115044 (DOI)

Funding

Korea Institute of Machinery and Materials
Basic Research Fund NK262C