Turkish Journal of Agriculture and Forestry
Abstract
This study focuses on enhancing the efficiency of agricultural decision support systems by integrating deep reinforcement learning techniques to optimize the path planning of agricultural vehicles. By decomposing the system into dynamics, action, strategy, reward, and discount factor components, a robust framework is established. A value-based reinforcement learning algorithm is introduced, enriched with policy gradient (PG) methodology to enable continuous and adaptive state-to-action decision-making process. A deep neural network (DNN) was used to approximate the state-action value function, resulting in the innovative DNN-PG-Q-learning algorithm. This algorithm leverages an empirical playback pool to enable efficient and intelligent path planning for agricultural vehicles. Experimental results demonstrated the effectiveness of the algorithm in both static and dynamic obstacle environments, with trajectory lengths ranging from 1.02 to 1.06 times the straight-line distance. During plowing operations, the algorithm achieves a trajectory length to straight-line distance ratio of 1.08 to 1.10, while maintaining nearly constant steering action, thereby enhancing the efficiency and stability of straight-line navigation.
Author ORCID Identifier
QINGSHAN BAI: 0009-0007-7015-2316
DOI
10.55730/1300-011X.3289
Keywords
Deep reinforcement learning, intelligent agricultural vehicles, path planning, reinforcement learning algorithms, deep neural networks
First Page
598
Last Page
611
Publisher
The Scientific and Technological Research Council of Türkiye (TÜBİTAK)
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
BAI, Q (2025). Deep reinforcement learning-driven decision support system design for precision agriculture: optimizing farming and resource use. Turkish Journal of Agriculture and Forestry 49 (3): 598-611. https://doi.org/10.55730/1300-011X.3289
