Hashim et al. (2025) Automated Rose Farming with IoT and Machine Learning: A Real-Time Predictive Irrigation System

Identification

• Journal: Advanced and Sustainable Technologies (ASET)
• Year: 2025
• Date: 2025-12-01
• Authors: Nur Zatil 'Ismah Hashim, Noramalina Abdullah, Intan Sorfina Zainal Abidin
• DOI: 10.58915/aset.v4i2.2707

Research Groups

School of Electrical and Electronic Engineering, Universiti Sains Malaysia, 14300 Nibong Tebal, Pulau Pinang, Malaysia.

Short Summary

This study developed an IoT-based automated rose farming system integrating machine learning for real-time environmental monitoring and intelligent irrigation control. The system achieved 100% classification accuracy for irrigation needs and demonstrated successful end-to-end operation, offering a cost-effective and scalable solution for smart floriculture.

Objective

• To develop a real-time, low-cost predictive irrigation system for rose cultivation that converts IoT sensor streams into pump-duration commands on a VisionFive2 RISC-V controller, closing the loop from sensing to actuation.
• To build an end-to-end platform for continuous monitoring and MQTT-based telemetry.
• To train and deploy a Random Forest regression model that predicts irrigation time from soil moisture variables.
• To validate closed-loop actuation and skip-logic behavior under controlled tests.

Study Configuration

• Spatial Scale: Indoor, controlled test environment for a single rose plant.
• Temporal Scale: Real-time continuous operation with a 5-minute waiting period after each irrigation cycle.

Methodology and Data

• Models used: Random Forest Regression model.
• Data sources:
  • Real-time sensor data from SHT3x (temperature and humidity) and resistive soil moisture sensors (digitized by MCP3008 ADC).
  • Manually constructed dataset for model training, comprising ambient temperature, humidity, current soil moisture, and a target soil moisture level (50%), with corresponding irrigation times.
  • Data transmitted via MQTT protocol for live monitoring.

Main Results

• The Random Forest model achieved 100% classification accuracy for differentiating between above- and below-median irrigation needs, with precision, recall, and F1-scores of 1.00 for both classes.
• The model's predictions were physically plausible, aligning with a heuristic of approximately 1 second of pump activation per 10% soil moisture increase and avoiding sub-second activations.
• End-to-end system validation confirmed accurate real-time data acquisition (e.g., 31.03 °C temperature, 66.57% humidity, 30.98% soil moisture), successful model inference (e.g., predicted 1.9 s irrigation time), precise actuation (0 s timing error on a 1.9 s command), and correct skip-logic for predictions below 1 second.
• Real-time telemetry to a public MQTT broker was successfully demonstrated, allowing live monitoring of sensor data and valve status.

Contributions

• Development of a cost-effective and scalable IoT-based automated irrigation system for rose cultivation, addressing the limitations of manual irrigation.
• Integration of real-time sensor data with a machine learning model (Random Forest) on an embedded RISC-V platform (VisionFive2) to predict optimal irrigation durations.
• Implementation of a closed-loop control system that maintains soil moisture around a target set-point, prevents unnecessary short pump cycling via a < 1 second guard, and provides live remote monitoring through MQTT.
• Validation of seamless sensing, inference, actuation, and cloud reporting across the entire control loop, demonstrating a practical solution for smart floriculture.

Funding

The School of Electrical and Electronic Engineering, Universiti Sains Malaysia, and StarFive Technology International Sdn Bhd.

Citation

@article{Hashim2025Automated,
  author = {Hashim, Nur Zatil 'Ismah and Abdullah, Noramalina and Abidin, Intan Sorfina Zainal},
  title = {Automated Rose Farming with IoT and Machine Learning: A Real-Time Predictive Irrigation System},
  journal = {Advanced and Sustainable Technologies (ASET)},
  year = {2025},
  doi = {10.58915/aset.v4i2.2707},
  url = {https://doi.org/10.58915/aset.v4i2.2707}
}