During food transportation, products are often subject to conditions which in no way promote the conservation of their biological properties. As a result of this loss of quality, there is a decrease in appreciation under the product. But more important than the loss of economic value is food waste. Remote monitoring systems, whose fundamental requirements relate to range and autonomy, make use of communications technologies to map characteristic crop parameters to reduce unnecessary application of resources or materials. This article proposes a monitoring system to record the conservation conditions inside refrigeration plants and vehicles, transmitting them via the internet of Things (IoT). This device is composed of an ARDUINO UNO Rev3 microcontroller that acquires the temperature and relative humidity every 5 minutes by means of a DHT 11 sensor, and uses the SIM800L module that provides real time communication data via GSM. It also incorporates a 3.7 V - 2600mAh battery which gives you an approximate 60 hour power range.

Deep Learning has been successfully applied to image recognition, speech recognition, and natural language processing in recent years. Therefore, there has been an incentive to apply it in other fields as well. The field of agriculture is one of the most important in which the application of artificial intelligence algorithms, and particularly, of deep learning needs to be explored, as it has a direct impact on human well-being. In particular, there is a need to explore how deep learning models for decision-making can be used as a tool for optimal planting, land use, yield improvement, production/disease/pest control, and other activities. The vast amount of data received from sensors in smart farms makes it possible to use deep learning as a model for decision-making in this field. In agriculture, no two environments are exactly alike, which makes testing, validating, and successfully implementing such technologies much more complex than in most other sectors. Recent scientific developments in the field of deep learning, applied to agriculture, are reviewed and some challenges and potential solutions using deep learning algorithms in agriculture are discussed. Higher performance in terms of accuracy and lower inference time can be achieved, and the models can be made useful in real-world applications. Finally, some opportunities for future research in this area are suggested. The ability of artificial neural networks, specifically Long Short-Term Memory (LSTM) and Bidirectional LSTM (BLSTM), to model daily reference evapotranspiration and soil water content is investigated. The application of these techniques to predict these parameters was tested for three sites in Portugal. A single-layer BLSTM with 512 nodes was selected. Bayesian optimization was used to determine the hyperparameters, such as learning rate, decay, batch size, and dropout size. The model achieved mean square error (MSE) values ranging from 0.07 to 0.27 (mm d–1)² for ETo (Reference Evapotranspiration) and 0.014 to 0.056 (m³m–3)² for SWC (Soil Water Content), with R2 values ranging from 0.96 to 0.98. A Convolutional Neural Network (CNN) model was added to the LSTM to investigate potential performance improvement. Performance dropped in all datasets due to the complexity of the model. The performance of the models was also compared with CNN, traditional machine learning algorithms Support Vector Regression, and Random Forest. LSTM achieved the best performance. Finally, the impact of the loss function on the performance of the proposed models was investigated. The model with the mean square error (MSE) as loss function performed better than the model with other loss functions. Afterwards, the capabilities of these models and their extension, BLSTM and Bidirectional Gated Recurrent Units (BGRU) to predict end-of-season yields are investigated. The models use historical data, including climate data, irrigation scheduling, and soil water content, to estimate endof- season yield. The application of this technique was tested for tomato and potato yields at a site in Portugal. The BLSTM network outperformed the GRU, the LSTM, and the BGRU networks on the validation dataset. The model was able to capture the nonlinear relationship between irrigation amount, climate data, and soil water content and predict yield with an MSE of 0.017 to 0.039 kg/ha. The performance of the BLSTM in the test was compared with the most commonly used deep learning method called CNN, and machine learning methods including a Multi-Layer Perceptrons model and Random Forest regression. The BLSTM out-performed the other models with a R2-score between 0.97 and 0.99. The results show that analyzing agricultural data with the LSTM model improves the performance of the model in terms of accuracy. The CNN model achieved the second-best performance. Therefore, the deep learning model has a remarkable ability to predict the yield at the end of the season. Additionally, a Deep Q-Network was trained for irrigation scheduling. The agent was trained to schedule irrigation for a tomato field in Portugal. Two LSTM models trained previously were used as the agent environment. One predicts the total water in the soil profile on the next day. The other one was employed to estimate the yield based on the environmental condition during a season and then measure the net return. The agent uses this information to decide the following irrigation amount. LSTM and CNN networks were used to estimate the Q-table during training. Unlike the LSTM model, the ANN and the CNN could not estimate the Qtable, and the agent’s reward decreased during training. The comparison of the performance of the model was done with fixed-base irrigation and threshold-based irrigation. The trained model increased productivity by 11% and decreased water consumption by 20% to 30% compared to the fixed method. Also, an on-policy model, Advantage Actor–Critic (A2C), was implemented to compare irrigation scheduling with Deep Q-Network for the same tomato crop. The results show that the on-policy model A2C reduced water consumption by 20% compared to Deep Q-Network with a slight change in the net reward. These models can be developed to be applied to other cultures with high importance in Portugal, such as fruit, cereals, and grapevines, which also have large water requirements. The models developed along this thesis can be re-evaluated and trained with historical data from other cultures with high production in Portugal, such as fruits, cereals, and grapes, which also have high water demand, to create a decision support and recommendation system that tells farmers when and how much to irrigate. This system helps farmers avoid wasting water without reducing productivity. This thesis aims to contribute to the future steps in the development of precision agriculture and agricultural robotics. The models developed in this thesis are relevant to support decision-making in agricultural activities, aimed at optimizing resources, reducing time and costs, and maximizing production.

This paper proposes a utilization of a reinforcement learning (RL) algorithm to control the packet transmission of multiple devices of a Cognitive Internet of Things (IoT) wireless communication system. The proposed approach consists of adopting a Markov chain to model the states of the communication system and its transitions, providing the required parameters to determine actions to the system using a Q-Learning algorithm. This paper also presents a performance evaluation of the developed algorithm in comparison to some scheduling algorithms in terms of: utility function, flow rate, buffer occupancy, packet loss rate, etc.

Industrial manufacturing companies are constantly faced with the challenge of increasing productivity, reducing rework costs and achieving rapid increases in production. The pressure, to keep production costs under control, is intense. Failure and poor quality in the production process is extremely dangerous to the competitiveness of manufacturing industries. According this context, it was purposed a Quantitative and a Qualitative case study at XGreen S/A assembly line of sprayers self-propelled. The study has got the proposal to verify the feasibility and the possibility to stablish a distinguished Quality in this product line, through the implementation of the intelligent tools: wireless torque wrenches, toolings (devices) and MES XGAAT (Manufacturing Execution System - X Green Assembly Assist Tool), which correspond to the materialization of Industry 4.0 concepts, within the scope of XGreen S/A. In short, it is expected through this study can itself to promote a reduction of Failure Per Machines (FPM), to improve the process traceability, the torque control as well as the Distinct Quality, in the assembly line of the sprayers from XGreen company. Although the implementation of the MES XGAAT system and the intelligent tools have been done into a conservative way, the quarterly reports from the Quality team suggest a considerable improvement regarding the Production indicators. The quarterly reports from Quality team have showed an improvement related to the Production´s indicators and a reduction of 27% in the Failures Per Machine (FPM), which was the main goal of this research project.

In an increasingly dynamic, complex and competitive market, proper maintenance strategy and maintenance actions are essential for companies to keep their processes and equipment operating safely and reliably, to meet quality requirements and lower overall costs. In this scenario, the implementation of machine condition monitoring solutions plays an important role, and this process is currently considered one of the main levers in global performance research. Inserted in this context, the present work aims to develop a low-cost prototype with the application of an ESP32 microcontroller, open-source software platforms and the concept of IoT (Internet of Things) to cover the functionality of data collection and monitoring of electrical quantities and environmental aspects of an electrical panel in continuous time. Additionally, the use of the Big Data concept for data processing is foreseen, and the construction of a model that generates alerts and provides subsidies for health prognosis, implementation of maintenance/performance indicators and failure prediction. This solution is expected to help companies implement a Condition Based Maintenance (CBM) strategy, make more assertive decisions in maintenance actions, reduce costs in corrective interventions and point out ways for asset management to be more efficient. At first, a prototype was developed with a total cost of US$ 155.99, tests and calibration of the sensors were carried out in the laboratory. Finally, the solution in an industrial environment was applied, in which the following variables were monitored: ambient temperature, relative air humidity, dew point and barometric pressure inside the panel, opening/closing status of the doors of the Motor Control Center (MCC), temperature and operating status of a circuit breaker and contactor, in addition to electrical currents and voltages of 2 motors of the primary wheat milling cylinder bank. The merit of the work can be attributed to the possibility of applying Industry 4.0 concepts to solve real problems and the high potential for scalability.

As a consequence of easy access to wireless technologies and the development of low-cost microcontrollers, the Internet of Things (IoT) has gained a lot of space in our globalized world. In this work, some tests will be exposed in order to show the operation of the wireless mesh network (Wireless Mesh Network - WMN) and its particularities for sensing. In addition, the possibility of applying WMNs to detect parking spaces in vehicles is presented, a scenario in which wireless nodes can be connected at distances ranging up to about 120 meters. In the proposed WMN research, wireless communication was carried out using ESP-8266 devices equipped with a Wi-Fi communication radio, where the WMN shared the same credentials (Service Set Identifier - SSID, access password and port (Transmission Control Protocol/Internet Communication Protocol - TCP/IP) to transfer data in ad-hoc type networks, in which there is a connection only between the WMN nodes, or in networks with external access, where there is a gateway node for interconnection of the WMN with the internet. With the implementation of this Course Completion Work - TCC, it was verified, in an experimental way, the viability of the proposal for i) the creation of sensor nodes to detect vacancies in vehicular parking lots; ii) the creation of WMN using sensors nodes and message forwarding nodes; iii) performance evaluation of the proposed WMN, through the generation of connection/disconnection scenarios of network nodes and verification of fault tolerance scenarios from the creation of redundant communication paths.

Internet of Things (IoT) devices, and Wireless Sensor Networks (WSNs), as pivotal elements in the actualisation of smart cities, require unlimited energy supply to enhance long lifespan and uninterruptible functionalities. Energy Harvesting (EH) techniques answer to many energy demands of perpetual devices, by converting various ambient environmental energy to usable electrical form for their sustenance. In this thesis, the review of various ambient energies available in the city that can be explored by different EH transduction mechanisms for the purpose of smart city automation was carried out. Thereafter, the PhD thesis explored the often windy weather of Aveiro city to investigate how to use ’home chimney pinwheels’ (HCP) on rooftops as an energy harvesting mechanism to convert energy from Wind to electrical source in a smart building in order to power IoT and other autonomous devices connected within such facility. The first step in the study was to justify reasons for which this means of harvesting Wind energy is considered smarter than the existing methods of using turbines of different capacities to harness Wind energy. This was achieved with the use of ’waterfall’ Project Management Methodology (PMM), whereby Wind energy delivery turbines that are grid connected at different capacities were categorised based on project management (PM). Hence, comparative analysis of different turbine capacities were made involving economic considerations; with delivery time, based on load target. The acronym ’Smart Turbine Energy Harvester (STEH)’ was therefore sustained for the use of HCP as turbines to deliver low amount of energy to IoT and other autonomous devices. Subsequently, the HCP was converted to an STEH by the use of off-the-shelve components, while the assembly part was fabricated with the aid of 3-D printer, and an electromagnetic converter by which means the Wind energy was harvested, was adapted from a brushless DC motor. This was mechanically fastened to the circular base of the 18-blade HCP. The HCP-STEH was tested for various Wind speed in the laboratory and experiments revealed that it can produce energy at very low Wind speeds. Experiments in simulated wind; in the laboratory yielded open circuit voltage of 0.3 ∼ 16 V, for wind speed between 0.6 ∼ 16 km/hr. In addition, the use of the smart turbine energy harvester as a smart building energy source was demonstrated on rooftop by connecting the set-up via a power management unit (PMU) to power ESP32 micro controller serving as IoT module across a 100 kΩ resistive load. Observation of the STEH output was made via cloud using LORA and the set-up was monitored on thingspeak cloud. On roof-top, with PMU connected to the STEH in ambient wind, a voltage of up to 4.1 V was generated; more so, maximum current of 6.6 mA and voltage of 9 V were recorded during cloud observation. This amount of energy is considered sufficient to operate low power IoT devices deployable around a smart city.

Sports practice is a common way people find to relieve the high stress levels. Sometimes it is impossible to move around and look for a place to do that, but the new technologies allow the practice of sports and recreational activities afar through the internet, in a comfortable way. The technological progress is responsible for the virtualization of many tasks. Nowadays, online multiplayer games try to be real, but they only focus on scenery virtualizations. As we know there already exists various UAV competitions, but sometimes, due to the many rules imposed by the aerial space, it isn't possible and safe to train when and where we want. To promote a better quality of life, it is being developed a platform with a totally distributed architecture based on the Internet of Drones concept, as it is shown in this article, that will allow the control of real unmanned aerial vehicles at anytime, from anywhere. The solution is composed by a high-scalability and a low-cost modeling system that will permit UAV users to practice their control skills in a comfortable and safe way. This interaction with a real vehicle will relieve the high stress levels caused by daily life.

The general increase in life expectancy and the consequent ageing of the general population impose major challenges to modern societies. Most elderly people experience the typical problems related to old age, such as chronic health problems, as well as sensory and cognitive impairments. In addition, in today’s modern societies, where families have less and less time to look after their older relatives, the isolation of the elderly is a real concern and a highly recurrent problem, which is enhanced when they live alone. To solve, or at least minimize, these problems, a smart home monitoring system was developed, as presented and described in this paper. This solution is implemented based on a sensory network which detects anomalous behaviors, immediately triggering a warning to the caregiver or family. A strong concern when developing a project of this kind is the physical security of the elderly. Houses tend to have hazardous objects and characteristics that may inflict serious injuries to their occupants or, in extrema, even death. As time goes by, the elderly start losing muscle mass and osteoporosis may appear, as well as vision and hearing impairments, which increase the likelihood of falling. Several other serious accidents may also occur, such as gas leaks, floods and fire outbreaks. Therefore, this population would strongly benefit from a solution which helps predict and even prevent accidents before they happen.

Daily activities have been increasingly supported by intelligent devices and applications. Smart devices are constantly communicating through the Internet of Things (IoT) networks, either by sending collected data and notifying the actions taken or by receiving instructions for actions to be taken. Most of this communication requires the confidentiality of data through the usage of encryption algorithms, being the Advanced Encryption Standard (AES) algorithm one of the most used. However, how do the operation modes of AES algorithm perform in a resource-constraint device? This paper aims to evaluate the impact on the time to encrypt and decrypt different sized messages in IoT devices when using each one of the five AES modes of operation and the three key sizes defined. The test scenario was implemented using two programming languages, running on a Raspberry Pi device. The results achieved infers that Python was quicker and had a more homogeneous result set than JavaScript implementation in most AES operation modes. These results help to understand the trade-off between IoT devices' security needs and delays in communication caused by the selection of the AES algorithm operation mode.