Forecasting Air Temperature on Edge Devices with Embedded AI.

Area of Science:

• Agricultural Technology
• Artificial Intelligence in Agriculture
• Environmental Monitoring

Background:

• Smart Agriculture integrates Internet of Things (IoT) and Machine Learning (ML) to enhance agricultural productivity and sustainability.
• Accurate environmental monitoring, such as internal air temperature, is crucial for optimizing greenhouse conditions.

Purpose of the Study:

• To design and evaluate a Neural Network (NN)-based model for predicting greenhouse internal air temperature.
• To assess the feasibility of deploying the prediction model on resource-constrained edge devices.

Main Methods:

• Investigated time series prediction using past and present air temperature data.
• Evaluated three Neural Network (NN) architectures: Long Short-Term Memory (LSTM), Recurrent Neural Networks (RNNs), and Artificial Neural Networks (ANNs).
• Assessed model performance using Root Mean Squared Error (RMSE), Mean Absolute Percentage Error (MAPE), and coefficient of determination (R2), alongside computational complexity (NetScore).

Main Results:

• The best models achieved RMSE values between 0.289–0.402°C and MAPE values between 0.87–1.04%.
• A coefficient of determination (R2) greater than 0.997 was consistently observed for the top models.
• An Artificial Neural Network (ANN)-based model demonstrated superior prediction accuracy with low computational and architectural complexity, suitable for edge deployment.

Conclusions:

• The developed ANN model effectively predicts greenhouse air temperature with high accuracy.
• The model's efficiency makes it suitable for deployment on edge devices in Smart Agriculture applications.
• This research contributes to advancing precision agriculture through intelligent environmental control.