Control Techniques for Vision-Based Autonomous Vehicles for Agricultural Applications: A Meta-analytic Review

The agriculture industry continues to expand, showing significant potential in meeting the escalating food demand. Renowned for its plentiful resources and substantial agricultural yield, India is progressively leveraging advancements in technology for mass production of superior quality goods. Notably, the integration of robotics has revolutionized farming practices, encompassing activities such as fruit picking, sheep herding, pruning, weeding, spraying, and other farming tasks. The move toward automation not only mitigates the reliance on human labor but also enhances the overall productivity of the agriculture sector. This paper conducts a detailed examination of the progress in the realm of vision-based autonomous vehicles (AV) tailored for agricultural applications. Previous research has employed a range of machine learning algorithms, from linear and logistic regression, decision trees, support vector machines (SVM), naive Bayes, K-nearest neighbor (KNN) algorithms, to K-means. Nowadays, hybrid algorithms are increasingly employed in AV designs, addressing the constraints of single algorithms while elevating their efficiency. Furthermore, ensemble techniques are being used for efficient handling and accurate prediction over vast data quantities. These techniques amalgamate the predictive prowess of numerous base estimators, boosting robustness, and are predominantly employed today utilizing Python and R codes.