Integration Of Neural Networks and Robust Control for an Agricultural Robotic Arm Using Crop-Based Data

• Andrinirina Fabien Ravelonahina

  Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

  Author

• Rabearivelo Apotheken Gericha

  PhD and Associate Professor at the Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

  Author

• Robinson Matio

  HDR (Accreditation to Supervise Research) at the Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

  Author

Data-driven robotic systems are critical in modern agriculture, enhancing crop yield through adaptive decision-making and visual recognition. While robust control ensures high-performance manipulation of robotic arms, it is limited in nonlinear and uncertain environments. This study presents a hybrid control framework integrating feedforward neural networks (NNs) with robust control using crop-specific datasets, improving adaptability and precision. Visual classification based on the Heaviside threshold method is employed to identify produce quality and color, demonstrated here for potato harvesting.

Experimental validation was performed in a greenhouse using a 6-DOF robotic arm, with sensors for soil moisture, temperature, and real-time image capture. Results show that the hybrid control system effectively classifies crops and adapts arm movements under varying environmental conditions, achieving both stability and adaptability. Limitations of the model, such as restricted generalization to extreme weather conditions, are discussed.

1. Andrinirina Fabien Ravelonahina, Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

   Doctoral student Cognitive Sciences and Applications, University of Antananarivo, Madagascar 

2. Rabearivelo Apotheken Gericha, PhD and Associate Professor at the Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

   Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar

3. Robinson Matio, HDR (Accreditation to Supervise Research) at the Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo, Madagascar.

   HDR (Accreditation to Supervise Research), Doctoral School of Science and Technology of Engineering and Innovation, Cognitive Sciences and Applications, University of Antananarivo

1. Mayer, C. L. (2005). Agriculture and sustainable development. [Book].

2. Nikolov, E., et al. (2005). Robust control. Technical University of Sofia.

3. Ghodbane, H., et al. (2010). Control of 6-DOF manipulator with obstacle avoidance. Courrier du Savoir.

4. Zakaria, H. M., & Hamza, B. (2018). Dynamic modeling of a PUMA 600 robot. University of Blida. [Unpublished manuscript/technical report].

5. Wang, K. (2016). Robust control for FSW welding [Doctoral dissertation, ENSAM].

6. Lamy, C. (2022). Using neural networks for energy transport modeling [Doctoral dissertation, University of Bordeaux].

7. Zerzaihi, H., & Zarour, F. (2020). Image recognition with CNNs [Master’s thesis, University of Jijel].

8. Goulet-Fortin, J. (2009). Potato yield modeling with neural networks [Master’s thesis, Laval University].

9. Rivals, I. (1995). Process modeling and vehicle control with neural networks [Doctoral dissertation, University Paris VI].

10. Boudin, A., & Recher, F. (2023). Mathematics of neural networks. [Book].

11. Boubrit, A. (2015). Seasonal variation of mesofauna under potato cultivation [Master’s thesis, Mouloud Mammeri University].

• PDF

All crop and environmental datasets used in this study are available upon reasonable request from the corresponding author.

Copyright (c) 2025 Author

This work is licensed under a Creative Commons Attribution 4.0 International License.