Pawar et al. (2025) Adaptive Precision Agriculture Through Iot And Reinforced Machine Learning (Q-Learning): A Sustainable Approach For Optimized Plant Growth

Identification

• Journal: International Journal of Advances in Signal and Image Sciences
• Year: 2025
• Date: 2025-12-10
• Authors: Yogesh B. Pawar, Pote Suraj Vishwanath, Vishal Ambaji Bogam
• DOI: 10.29284/ijasis.11.6s.2025.9-22

Research Groups

• Department of Computer Science & Engineering, Asian International University, Manipur, India
• Department of Computer Engineering, MIT Academy of Engineering, Alandi, Pune, India
• School of Computing, MIT ADT UNIVERSITY, Pune, India

Short Summary

This study integrates IoT-based sensing with Q-learning reinforcement learning to optimize plant growth and resource efficiency in both controlled and natural field conditions. The findings validate IoT-based adaptive monitoring as a practical tool for improving crop yields, minimizing input use, and promoting long-term agricultural sustainability.

Objective

• To integrate IoT-based sensing with Q-learning reinforcement learning to optimize plant growth and resource efficiency in both controlled and natural field conditions, aiming to derive actionable insights for sustainable agriculture.

Study Configuration

• Spatial Scale: Controlled environments (greenhouses, simulated ideal growth settings) and natural field conditions (on-field trials, open fields).
• Temporal Scale: Real-time, continuous monitoring and iterative learning for dynamic adjustments over time.

Methodology and Data

• Models used: Q-learning algorithm (Reinforcement Learning), formalized as a Markov Decision Process (MDP). Multi-objective optimization for balancing growth and resource efficiency.
• Data sources: IoT-enabled sensors capturing real-time data on soil moisture (Sm(t)), soil temperature (ST(t)), air humidity (H(t)), air temperature (Ta(t)), light intensity (L(t)), and plant growth index (G(t)).

Main Results

• Controlled environments consistently outperformed open fields, achieving:
  • Greater average plant height (45.2 cm vs. 38.7 cm).
  • Larger leaf area index (122.4 cm²/leaf vs. 98.6 cm²/leaf).
  • Higher biomass yield (315 g vs. 260 g).
  • Superior growth consistency (92.3% vs. 78.4%).
  • Higher resource use efficiency (rated high vs. moderate).
• Q-learning models demonstrated faster convergence and higher cumulative rewards in controlled settings (152.4 cumulative reward, 450 episodes for convergence) compared to uncontrolled environments (121.7 cumulative reward, 620 episodes for convergence).
• The system effectively adapted to environmental variability in open fields, underscoring the robustness of reinforcement learning in real-world agriculture.
• Sustainability analysis showed IoT-RL systems outperformed traditional manual practices by up to 30% in water efficiency and achieved higher sustainability scores.

Contributions

• Proposes and validates an integrated IoT and Q-learning reinforcement learning framework for adaptive plant growth optimization.
• Demonstrates the effectiveness of the framework in enhancing plant growth and resource efficiency across both controlled and uncontrolled agricultural environments.
• Offers a scalable framework for improving crop yields, minimizing input use (water, fertilizer, energy), and promoting long-term sustainability in agriculture.
• Addresses critical gaps in scalability, adaptability, and accessibility within smart agriculture by providing a closed-loop system for intelligent, real-time decision-making.

Funding

Not specified in the paper.

Citation

@article{Pawar2025Adaptive,
  author = {Pawar, Yogesh B. and Vishwanath, Pote Suraj and Bogam, Vishal Ambaji},
  title = {Adaptive Precision Agriculture Through Iot And Reinforced Machine Learning (Q-Learning): A Sustainable Approach For Optimized Plant Growth},
  journal = {International Journal of Advances in Signal and Image Sciences},
  year = {2025},
  doi = {10.29284/ijasis.11.6s.2025.9-22},
  url = {https://doi.org/10.29284/ijasis.11.6s.2025.9-22}
}