Agricultural crop yield prediction is crucial for enhancing food security, optimizing resource use, and ensuring sustainable agricultural practices. This project focuses on enhancing food security and sustainable agricultural practices by predicting crop yields using machine learning techniques. This project leverages advanced data analysis techniques, including machine learning and statistical models, to accurately forecast crop yields. The study investigates the correlation between crop yield and crucial input variables such as nitrogen, phosphorous, potassium, rainfall, temperature, and fertilizer application. The primary objective is to develop accurate and reliable predictive models that enable farmers and agricultural stakeholders to anticipate crop yields, aiding in better planning and resource allocation. The research examines the correlation between crop yield and environmental factors such as nitrogen, phosphorus, potassium, rainfall, temperature, and fertilizer application. Multilinear Regression (MLR), Radial Basis Function (RBF), and Support Vector Machine (SVM) models are applied to predict yields, with SVM achieving the highest accuracy at 92.03%, followed by MLR at 88.56% and RBF at 75.36%. The data collection for this study includes pesticides usage, historical weather parameters, and fertilizer usage from the Peddapalli district, Telangana, India. MLR identifies linear relationships, RBF captures non-linear patterns, and SVM handles high-dimensional data to enhance prediction accuracy. The results indicate that while MLR and RBF provide valuable insights, SVM is the most robust tool for forecasting crop yields. This research holds significant potential for improving agricultural productivity and resource management, offering farmers crucial insights for better planning and allocation of resources.

To make it possible for computer vision to self-train and comprehend visual input, pattern recognition algorithms are mainly used. Advanced measurements are required every time for the early detection of armed threats because of decreasing accidents and terrorist attacks. Weapon detection systems are mostly used in public spaces such as stadiums, airports, key squares, and battlefields, whether they are in urban or rural settings for better security objectives. Based on cloud architecture, DL, and ML algorithms are used by contemporary closed-circuit television surveillance and control systems to detect weapons. Using the Raspberry Pi as an edge device and the Efficient model to construct the weapons detection system, edge computing can be used to address these problems. The text report, including the image processing results, is sent to the cloud platform so that the operator can review it further. Soldiers can outfit themselves with the recommended edge node, headphones, and augmented reality glasses for visual data output to receive alerts about armed threats. Furthermore, we can improve our method's performance by adding more training data and changing the network architecture. The primary object of this paper is to build a model for detecting weapons such as pistols and rifles. The data will be taken from the Kaggle dataset. Our results and recommendations will help new researchers and related organisations.

In recent years, the adoption of wireless sensor systems has significantly grown in many different kind of applications, like environmental monitoring, chemical analysis, food safety, health monitoring and quality analysis in industrial environments, and contributed to the development of the Internet of Things (IoT) paradigm.
Typically, sensors are interfaced with computing devices (e.g., microcontrollers, FPGAs, etc.) that acquire sensor data using an analog-to-digital converter (ADC). Since individual sensor nodes are usually powered by batteries, power consumption is a critical aspect that significantly impact the sensor node lifetime. In order to reduce power consumption, sensors can be interfaced with computing devices by using the sensor-to-microcontroller direct interface (SMDI), without the need to use an ADC which requires higher power. The SMDI exploits Schmitt trigger circuits that are typically integrated in the general purpose input output (GPIO) interface of a microcontroller. SMDI can be applied to many different kind of sensors, such as resistive and capacitive sensors, as well as any other sensors producing analog output voltage, and can allow sensor measurements with lower cost and power consumption than the traditional ADC based data acquisition.
In this study, we investigate the application of SMDI technique when it is employed to acquire data from a non-linear negative temperature coefficient (NTC) thermistor. We evaluated the accuracy of temperature measurements by means of electrical level simulations, considering real operating conditions and two well know models from literature (Steinhart-Hart model and polynomial model) to estimate the accuracy of temperature measurements. The results have shown that the temperature estimation using data obtained by SMDI measurements provides good accuracy. In particular, the Steinhart-Hart model provides more accurate results (average error 0.078 °C) than the polynomial model, that features an average error of 0.28 °C.

The growing world population necessitates more efficient food production, particularly in agriculture. The traditional irrigation techniques usually result in overwatering or underwatering, which wastes energy and water and reduces agricultural productivity. The smart agriculture optimizes food production, resource management, and labor. This study introduces an intelligent irrigation and monitoring system that uses the Internet of Things (IoT) to automate water pump management and monitor sun light, temperature, and humidity levels without human interaction. The system's hardware components include a soil moisture sensor, sun light sensor, temperature and humidity (DHT11) sensor, ESP32 microcontroller, and pump motor. The sensors are in charge of gathering the information that the ESP32 microcontroller needs in order to properly operate the pump motor.To operate and monitor data from the sensors remotely, the ESP32 is also integrated with the well-known Ubidots server via a wireless sensor network. Initially, sensors such as DHT11, soil moisture, and sunlight level collect data from the field and send it to the ESP32 microcontroller. The microcontroller then compares the received data to the previously stored data. If the values are greater than the threshold, the associated devices turn on and update the sensor value and pump motor condition to the Ubidots server.

Electrochemical Impedance Spectroscopy (EIS) is a widely used technique to analyse the properties of electrode surfaces and bulk electrolytes in electrochemical sensors. At low frequencies (below 100 Hz), EIS provides insights into the electrode-electrolyte interface, particularly the double-layer capacitance. At higher frequencies (>1 kHz), the double-layer capacitance effects diminish, and the impedance is predominantly influenced by electrolyte resistance. Designing effective EIS electrodes requires a comprehensive understanding of both the electrode-electrolyte interface and bulk electrolyte properties. In this study, we developed a COMSOL-based time-dependent model to simulate and understand the EIS response of a planar two-electrode sensor. The intended application is to simulate the EIS response in a buffer solution used for biological cell perfusion in organ-on-chip systems.

Simplified models, such as the Randles circuit for parallel plate capacitors, offer basic estimations but become inadequate for complex geometries like planar interdigitated electrodes and electrolytes with bulk reactions. Numerical simulations offer a more accurate approach in these cases. While traditional analytical models provide steady-state analysis, time-dependent numerical models offer detailed insights into the dynamic processes at the electrode-electrolyte interface during electrode excitation.

Our COMSOL model captures the dynamic processes in the Debye layer, providing a comprehensive understanding of electrode-electrolyte interactions and transient behaviours. It employs the Nernst-Planck and Poisson equations to solve for ion concentration and electric potential, respectively. It includes practical parameters such as electrode sheet resistance and parasitic capacitances, enhancing its representational accuracy. The EIS responses from the model were validated against experimental results at various NaCl concentrations (1, 10, and 100 mM). The model results show a comparable impedance spectrum and values to experimental data, particularly at higher frequencies (>1 kHz), demonstrating its effectiveness in capturing the electrochemical behaviour of the system.

Amyotrophic Lateral Sclerosis (ALS) is a relentlessly progressing neurological disease with limited treatment options. The advent of extensive global datasets and advanced machine learning models offers new opportunities to evaluate potential prognostic, inspection, and diagnostic indicators. Additionally, emerging categorization and staging systems aim to accurately stratify patients into distinct prognostic categories. This study employs an array of machine learning algorithms to predict ALS diagnoses, including decision trees, ensemble methods, gradient boosting algorithms, and support vector machines. Specifically, it uses classifiers like DecisionTree, ExtraTree, Random Forest, Extra-Trees, XGB, LGBM, CatBoost, AdaBoost, SVC, and MLPClassifier. These models range from basic tree-based methods, which split data based on feature values for predictions, to advanced ensemble techniques like Random Forests and gradient boosting, which combine several models to enhance accuracy and robustness. The Support Vector Machine (SVC) identifies the optimal hyperplane to separate classes, while the MLPClassifier, a type of neural network, captures complex data patterns. This diverse approach leverages the unique strengths of each algorithm, providing a comprehensive evaluation of model performance for ALS diagnosis. Results show that the CatBoost classifier achieved the highest performance, with an accuracy of 0.85 and an AUC of 0.97. Other significant models include XGB, RandomForest, and ExtraTrees classifiers, each showing an accuracy of around 0.75 but with varying AUC values.

In classification tasks, custom sensors have traditionally been employed to achieve accuracy scores. While numerous studies have reported high accuracy rates, there has been limited discussion on real-time predictions or practical applications in most research papers. Real-time prediction of object material properties is crucial for enhancing the tactile sensing capabilities of robotics in industrial settings. This study proposes the use of Commercial Off-The-Shelf (COTS) tactile sensors for hardness classification, utilizing small datasets for model training and real-time prediction. Testing involves evaluating the ability of robotic grippers to accurately predict the hardness of new, unknown objects, categorizing them into two classes (soft, hard) or three classes (hard, soft, flexible). Results obtained from a multiple-algorithm approach reveal an 80% accuracy rate for binary classification, with real-time tests demonstrating 2 out of 3 correct predictions for most sensors. For ternary classification, the accuracy rate is 70%, with 2 out of 3 correct predictions from at least one sensor. These findings highlight the capability of COTS sensors to perform real-time hardness classification effectively. This also highlights that COTS sensors have capabilities and flexibility based on their dimensional architecture that they can be used in many different robotics applications without investing time in the development of a specific use-case sensor for classification task within robotic tactile sensing.

In the following work, a generalizable approach for robotically aided quality inspection is proposed and applied in the field of plastic injection molding of small parts. While artificial intelligence has shown promising results for quality inspection, it often requires large training datasets, which are impractical for industrial applications. Defective, or anomalous parts are heavily underrepresented in normal production in comparison to expected parts which makes it a poorly posed problem for supervised learning approaches. In the proposed concept of this work, this issue is tackled by combining an automated inspection procedure with an anomaly detection approach for defect detection. A 7-DoF robotic manipulator was used to automate the part handling in front of an industrial optical camera sensor. Sampling was done on multiple days with varying surrounding conditions to ensure a heterogenous dataset. The captured images were used to train a PaDiM anomaly detection network to reconstruct a normal image of the part. To determine the capabilities of the developed model, different aspects were evaluated: the amount of training data, the output heatmap resolution and the anomaly decision approach. The results for the best combination of parameters show that various defects, such as particles, scratches, missing structures and deformations can be detected with defect detection rates up to 100% while maintaining approximately 91% true negative rates using a small dataset, consisting of 117 parts and low-resolution anomaly heatmaps to enable faster processing times. Furthermore, the study also found that the anomaly decision approach had minimal impact on prediction quality, whereas reducing training samples below 100 and heatmap resolutions below 1024x1024 significantly decreased prediction accuracy. The combination of robotic automation and anomaly detection is thus suitable for quick adaptation to multiple use cases.

The selection of best gripper for pick and place robot arm, is always a tedious task to choose from various grippers available. Most of the gripper does not comply with pay load capacity of robot arm and also lacks in efficient handling of objects. This paper assists in solving the issue i.e. how to choose an adequate gripper for pick and place operation considering its object handling property. Therefore, performance testing of parallel and 3 finger gripper are done under simulation environment based on human hand grasping taxonomy. Grasping force and number of contact points between gripper and objects are parameters for comparative study. Simulation experiments are conducted by gazebo simulator running under ROS, to pick and place 10 objects. Based on human hand grasping taxonomy simulation experiments are divided into 3 categories such as normal, misaligned and rotation grasping. Points 0 and 1 are awarded to grippers i.e. 1 is for easy grasping and 0 for difficult grasping or no grasping. From experiments, overall grasping score of parallel and 3 finger gripper are 0.8779 and 0.9667 respectively which leads to conclusion that not only grasping of 3 finger gripper is superior as compared to parallel gripper but also it performs very well on 3 grasping taxonomies of human hand and able to handle fragile and deformable objects efficiently. By addressing the objectives of proposed study, researcher can easily select gripper for a particular application and also can develop more capable, adaptable, and cost-effective manipulation systems.

Zinc oxide is a multifunctional material, and composite nanomaterials based on it are used in optoelectronic devices and solar cells. P-type oxides CuO is added to the n-type semiconductor ZnO to create photosensitive resistors. In this work, nanocrystalline composite CuO-ZnO thin films were synthesized using solid-phase pyrolysis by the two-step method. Zn(CH₃COO)₂∙2H₂O, Cu(CH₃COO)₂∙2H₂O and C₁₉H₂₉COOH were used as the precursors. At the first stage, an intermediate product was obtained, which is a mixture of zinc and copper salts. Then, a solution of this intermediate product in an organic solvent was applied to pre-prepared substrates. After calcination for 2 hours at 600 °C, a film coating was formed. The molar ratio of Cu:Zn is 1:99 and 5:99. By XRD, the obtained materials are polycrystalline and two-phase. The peak intensity of ZnO is significantly more pronounced compared to CuO, which appears as the additive content increases. No other phases were detected. SEM studies have revealed that the films are homogeneous. The average size of spherical particles is 18 nm. n-p heterojunction formed between ZnO and CuO nanocrystallites in contact with each other. Optical measurements made it possible to determine the energy level of the band gap in the obtained composites. This energy level was found to be lower than that of ZnO, which explains why the composites are sensitive not only to UV light but also to visible light. The resulting materials are sensitive not only to UV light, but also to visible radiation, suggesting that they can be used to create photosensitive resistive sensors.

This research was financially supported by the Russian Science Foundation 24-29-00203 (https://rscf.ru/project/24-29-00203/) at the Southern Federal University.