Abstract
Pollination is critical for crop development, especially those essential for subsistence. This study addresses the pollination challenges faced by Actinidia, a dioecious plant characterized by female and male flowers on separate plants. Despite the high protein content of pollen, the absence of nectar in kiwifruit flowers poses difficulties in attracting pollinators. Consequently, there is a growing interest in using artificial intelligence and robotic solutions to enable pollination even in unfavourable conditions. These robotic solutions must be able to accurately detect flowers and discern their genders for precise pollination operations. Specifically, upon identifying female Actinidia flowers, the robotic system should approach the stigma to release pollen, while male Actinidia flowers should target the anthers to collect pollen. We identified two primary research gaps: (1) the lack of gender-based flower detection methods and (2) the underutilisation of contemporary deep learning models in this domain. To address these gaps, we evaluated the performance of four pretrained models (YOLOv8, YOLOv5, RT-DETR and DETR) in detecting and determining the gender of Actinidia flowers. We outlined a comprehensive methodology and developed a dataset of manually annotated flowers categorized into two classes based on gender. Our evaluation utilised k-fold cross-validation to rigorously test model performance across diverse subsets of the dataset, addressing the limitations of conventional data splitting methods. DETR provided the most balanced overall performance, achieving precision, recall, F1 score and mAP of 89%, 97%, 93% and 94%, respectively, highlighting its robustness in managing complex detection tasks under varying conditions. These findings underscore the potential of deep learning models for effective gender-specific detection of Actinidia flowers, paving the way for advanced robotic pollination systems.

Abstract
Pruning consists on an agricultural trimming procedure that is crucial in some species of plants to promote healthy growth and increased yield. Generally, this task is done through manual labour, which is costly, physically demanding, and potentially dangerous for the worker. Robotic pruning is an automated alternative approach to manual labour on this task. This approach focuses on selective pruning and requires the existence of an end-effector capable of detecting and cutting the correct point on the branch to achieve efficient pruning. This paper reviews and analyses different end-effectors used in robotic pruning, which helped to understand the advantages and limitations of the different techniques used and, subsequently, clarified the work required to enable autonomous pruning. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

Abstract
This paper introduces a deep learning approach for detecting plant leaf diseases. The objective is to develop robust models capable of accurately identifying plant diseases across various image backgrounds, thereby overcoming the limitations of existing methods that often rely on controlled laboratory conditions. To achieve this, a combination of the PlantDoc dataset and Web-sourced data of plant images from online platforms was used. This paper implemented and compared state-of-the-art *cnn architectures, including EfficientNet-B0, EfficientNet-B3, ResNet50, and DenseNet201, all fine-tuned specifically for leaf disease classification. A significant contribution is the application of enhanced data augmentation techniques, such as adding Gaussian noise, to improve model generalisation. Results indicated varied performance across the datasets, with EfficientNet models generally outperforming others. When trained and tested on the PlantDoc dataset, EfficientNet-B3 achieved the highest accuracy of 73.31%. In cross-dataset evaluation, EfficientNet-B3 reached 76.77% accuracy when trained on PlantDoc and tested on the Web-sourced dataset. The best performance occurred when training on the combined dataset and testing on the Web-sourced data, resulting in an accuracy of 80.19%. Class-wise F1-scores revealed consistently high performance (>0.90) for diseases such as apple rust leaf and grape leaf across models. This paper contributes to the comparative analysis of various datasets and model architectures for effective leaf disease detection

Abstract
A novel non-destructive analytical method for early diagnosis of two bacterial diseases, Pseudomonas syringae and Xanthomonas euvesicatoria, in tomato plants, using ultraviolet-visible (UV-Vis) transmittance spectroscopy and chemometric models, is developed. Plant-pathogen interactions caused tissue damage that generated non-linear data patterns compared to the control set (healthy samples), which challenges traditional discrimination models, even when employing non-linear discriminant approaches. Alternatively, an authentication task to conduct oneclass classification relying on a data-driven version of soft independent modeling of class analogy (DD-SIMCA) is a wise choice due to its quadratic approach, proper to deal with non-linear data. DD-SIMCA detached the target class (control healthy plant leaflet tissues) from all other samples (target class and non-target class of plant leaflet tissues inoculated with two bacteria, even before the manifestation of macroscopic lesions associated with the diseases) by capturing the main similarities within the samples of the target class through the full distance that acts as a classification analytical signal, reaching 100 % sensitivity in the training and validation sets. Multivariate curve resolution - alternating least-squares (MCR-ALS) constrained analysis allowed the description of the bacterial inoculation process on diseased tissues through pure spectral signatures. DD-SIMCA results indicate that non-target class of samples with higher proximity to the acceptance boundary suggested that they were at earlier stages of infection when compared to more distant ones, presenting lower full distance values. These findings reveal that a handheld UV-Vis transmittance spectrometer is sufficiently sensitive to be used in acquiring biological data with suitable chemometric models for early disease diagnosis and prompt intervention.

Abstract
Fertilization is paramount for agriculture productivity and food security. Plant nutrition pre-established recipes and nutrient uptake are rarely managed by changing the fertilizer composition at the different stages of the plant life cycle. Herein we perform a literature review analysis - since the year 2000 and onwards - of the state-of-the-art capabilities of Nitrogen, Phosphorous, and Potassium (NPK) sensors for liquid fertilizers ( e.g. , hydroponics). From the initial search hits of 1660 results, only 53 publications had relevant information for this topic; from these, only 9 had NPK quantitative information. Qualitative analysis was performed by determining the number of publications for each nutrient, according to sample complexity and existing single, multiplexed or hybrid technologies. Quantitative assessment was performed by extracting the bias and linearity, the limit of detection and concentration ranges of sensor operation, framed into the context of the sensor technology development stage and sample compositional complexity. The most common technologies are colorimetry, ionselective electrodes, optrodes, chemosensors, and optical spectroscopy. The most abundant technologies are for nitrate quantification, from which ion-selective electrodes are the most widely used technology, and sensors for phosphate quantification are the less developed. Most are at low technological levels of development, not dealing with the complexity of agriculture samples due to matrix effects and interference. Measuring the fertilizer composition, nutrient uptake, the state of the chemical network, and controlling the release of nutrients using new functional materials, is one of the most important challenges ahead for the existence of precision fertilization. Intelligent sensing and smart materials are today the most successful strategy for dealing with matrix effects and interferences, being led by ion-selective electrodes and spectroscopy technologies.

Abstract
Job shop scheduling problems are common in the engineering field. In spite of some approaches consider just the most important objective to optimize, several other conflicting criteria are also important. Multi-objective optimization algorithms can be used to solve these problems optimizing, simultaneously, two or more objectives. However, when the number of objectives increases, the problems become more challenging. This paper presents the results of the optimization of a set of job shop scheduling with unrelated parallel machines and sequence-dependent setup times, using the NSGA-III. Several instances with different sizes in terms of number of jobs and machines are considered. The goal is to assign jobs to machines in order to simultaneously minimize the maximum job completion time (makespan), the average job completion time and the standard deviation of the job completion time. These results are analysed and confirm the validity and highlight the advantages of this approach.