Reinforcement learning model in automated greenhouse control
Wedi'i Gadw mewn:
| Awduron: | , , , , |
|---|---|
| Fformat: | comunicación de congreso |
| Dyddiad Cyhoeddi: | 2023 |
| Disgrifiad: | Automated systems, controlled with programmed reactive rules and set-point values for feedback regulation, require supervision and adjustment by experienced technicians. These technicians must be familiar with the scenario where the controlled processes are carried out. In automated greenhouses, achieving optimal environmental values requires the expertise of a specialist technician. This introduces the need for an expert in the installation and the problem of depending on them. To reduce these inconveniences, the integration of three paradigms is proposed: user-centered design, deployment of data capture technology based on IoT protocols, and a reinforcement learning model. The objective of the reinforcement learning model is to make decisions in the programming of set-points for the climate control of a greenhouse. In this way, the need for manual and repetitive supervision of the specialized technician is reduced; meanwhile, the control is optimized. The design, led by an expert technician in greenhouse installations, provides the necessary knowledge to transfer to a reinforcement learning model. On the other hand, deploying the required set of sensors and access to external data sources increases the capacity of the learning model to be deployed to current installations. The proposed system was tested in automated greenhouse facilities under the supervision of a specialized technician, validating the usefulness of the proposed system. |
| Gwlad: | Kérwá |
| Sefydliad: | Universidad de Costa Rica |
| Repositorio: | Kérwá |
| Iaith: | Inglés |
| OAI Identifier: | oai:kerwa.ucr.ac.cr:10669/102269 |
| Mynediad Ar-lein: | https://hdl.handle.net/10669/102269 https://doi.org/10.1007/978-3-031-48642-5_1 |
| Allweddair: | reinforcement learning smart greenhouse Q-Learning |