Vertical farming can address the growing challenges of food scarcity, limited arable land, and environmental constraints in urban areas, which are shaped by rising populations and climate change. When implemented correctly, this method has the potential to eliminate the extensive use of arable land. Some key benefits include year-round crop production with minimal reliance on weather, reduced pesticide dependence, and efficient water use through closed-loop systems. 

In a paper presented at the 2025 IEEE European Technology and Engineering Management Summit, researchers describe a comprehensive framework for engineering smart agricultural systems that leverage the Internet of Things (IoT) and Artificial Intelligence (AI) technologies for sustainable urban farming. 

Urban Vertical Farming 

Urban vertical farming (UVF) is an agricultural approach that addresses challenges in traditional farming, including limited land, scarce water, and climate change. UVF employs hydroponic and aeroponic farming systems, which allow crops to be grown in a greenhouse-like structure and provide for effective space management with minimal weather impact and reduced water use. 

The UVF method enables sustainable farming by leveraging real-time monitoring, data analysis, and resource management to increase crop yields, reduce environmental damage, and support sustainable agriculture. 

This paper aims to identify the gaps in urban agriculture. It proposes a solution that emphasizes the use of IoT and AI technologies to increase crop production, improve management, enhance resource efficiency, and promote sustainability. All practices are aimed at protecting nature, mitigating the impacts of climate change, and improving public health. 

In the paper, the researchers navigate related work, focusing on smart agriculture and vertical farming in controlled environments, and on AI-powered precision farming, before introducing their proposed method.

Methodology & Results

The authors outline the methodological framework for the study’s short-term temperature forecasting in a vertical farm setting. 

Specifically, they describe how the raw IoT data were collected and processed, and how they were used to train and evaluate two distinct predictive approaches: a classical ARIMA model and a neural network-based LSTM model. The paper covers Experimental Site and Growth Conditions, Data Collection & Analysis, and Model Training, before discussing testing results.

Data collection infrastructure

 

Data was collected continuously at one-minute intervals for two weeks using both cloud and local systems to ensure robust, scalable data acquisition across a variety of use cases. Afterwards, short-term temperature forecasting enabling proactive climate control to optimize resource use and crop yield was done with classical time series approaches (ARIMA) and deep learning techniques (LSTM) based optimize resource use and crop yield. 

Next-hour temperature forecast.

 

The LSTM models appeared to track more closely during rapid changes, indicating that deep learning’s ability to perform sequence modelling may be helpful in instances of abrupt temperature changes. That small difference in error can be very helpful in a production setting, especially where tighter control of the climate can save energy and enhance the stability of crop growth.

Conclusion

The findings validate the use of data-driven approaches for forecasting to improve resource-efficient, high-output indoor agricultural systems. In the comparison, both techniques captured the most important changes in the data, indicating a relatively well-controlled environment. The LSTM approach achieved slightly better results when using sequential patterns over a 12-hour period, indicating that deep learning models that capture temporal relationships can improve performance. 

Such a forecast can enable the vertical farm’s management system to optimize cooling or heating in advance, ensuring ideal conditions for plants and improving energy efficiency. While ARIMA may perform well in relatively stable environments, the LSTM approach’s ability to capture sudden changes more effectively suggests it is superior in more volatile environments.

This study is limited in scope; future work can build on this research by incorporating additional features, such as humidity, CO2 concentration, and light, to improve predictive capabilities and develop a more effective control strategy. 

Interested in learning more about Smart Agriculture? The IEEE Xplore Digital Library offers over 14,000 publications on Smart Agriculture.

Interested in acquiring full-text access to this collection for your entire organization? Request a free demo and trial subscription for your organization.