Este projeto tem como objetivo a especicação de um enlace de comunicação digital para veículos aéreos não tripulados. Os principais desaos presentes no meio de transmissão serão evidenciados, assim como o impacto acarretado no sistema de comunicação. O projeto foi dividido entre a parte analógica e digital. A parte analógica tratará dos requisitos de potência para o devido funcionamento do canal através do procedimento de link budget. O projeto da parte digital, por sua vez, empregará a técnica de transmissão OFDM. No presente trabalho foi sugerido um método de estimação do canal utilizando os tons pilotos. O desempenho desta abordagem será medido através de uma simulação de monte Carlo / This project aims to design a digital data communication link for unmaned aerial vehicles. We will focus the main challenges and their impacts in the communication system. The project was divided in an analog and digital block. The analog block address the power requirements that make the system works by using a link budget procedure. The digital block will use the OFDM transmission technique. In this work we also suggest a channel estimation procedure via pilot tones. The performance of this approach will be measured by Monte Carlo Simulation

<div>In 21st century, interoperability in remote locations has always been a matter of contention. Interoperability is very closely related to internet and an efficient process saves a lot of time and money. With the advent of Wireless Sensor Networks (WSN), Native Internet Protocol (NIP) is considered as one of the most pragmatic solutions in market to address interoperability challenges and is gaining more attention in research. However, challenges like reliability, security of data, power consumption, range and maintenance, and accessibility of such internet in remote locations still remain a matter of concern, creating further barriers for interoperability. This research aims at proposing a viable solution to interoperability issues at remote locations, irrespective of its network or payload size, by integrating more advanced Wireless Sensor Protocols like Thread Protocol with a proposed Over The Air (OTA) file transfer functionality, into UAVs. Furthermore, this study analyzes power consumption, reliability, latency and scope of the proposed system and their applications in health care and industries.</div>

This dissertation investigates vulnerabilities in unmanned vehicles and how to successfully detect and counteract them. As we entrust unmanned vehicles with more responsibilities (e.g. fire-fighting, search and rescue, package delivery), it is crucial to ensure their safe operation. These systems often have not been designed to protect against an intelligent attacker or considering all possible interactions between the physical dynamics and the internal logic. Robust control strategies can verify that the system behaves normally under bounded disturbances, and formal verification methods can check that the system logic operates normally under ideal conditions. However, critical vulnerabilities exist in the intersection of these fields that are addressed in this work. Due to the complex nature of this interaction, only trivial examples have previously been pursued. This work focuses on efficient real-time methods for verification and validation of unmanned vehicles under disturbances and cyberattacks. The efficiency of the verification and validation algorithm is necessary to run it onboard an unmanned vehicle, where it can be used for self diagnosis. We begin with simple linear systems and step to more complex examples with non-linearities. During this progression, new methods are developed to cope with the challenges introduced. We also address how to counter the threat of unmanned aerial systems (UASs) under hostile control by developing and testing an estimation and control scheme for an air-to-air counter UAS system.<br>

Wild leek (Allium tricoccum) is a spring ephemeral of northeastern North America. In the Canadian province of Quebec, it is listed as threatened due to human harvesting, and in Gatineau Park its presence is used as an indicator of human impact. Wild leek grows in patches on the forest floor, and before the tree canopy develops its green leaves are clearly visible through the bare branches of deciduous forests, allowing it to be observed with optical remote sensing. This study developed and tested a new method for monitoring wild leek across large geographic areas by integrating field observations, UAV video, and satellite imagery. Three-cm resolution orthomosaics were generated for five <0.1 km2 sites from the UAV video using Structure-from-Motion, segmented, and classified into wild leek (WL) or other (OT) surface types using a simple greenness threshold. The resulting maps, validated using the field observations, had a high overall accuracy (F1-scores between 0.64 to 0.94). These maps were then used to calibrate a linear model predicting the per-pixel percentage cover of wild leek (%WL) from NDVI in the satellite imagery. The linear model calibrated for a Sentinel-2 image from 2018, covering all of Gatineau Park (~361 km2), allowed %WL to be predicted with an RMSE of 10.32. A similar model calibrated for a WorldView-2 image from 2018 was noisy (RMSE = 37.64), though much improved by resampling this image to match the spatial resolution of Sentinel-2, due to MAUP scale effect (RMSE = 13.06). Testing the potential for satellite-based monitoring of wild leek, the %WL prediction errors were similar when a new linear model was developed using the Sentinel-2 image from 2017 (RMSE = 12.84) and when the model calibrated with the 2018 Sentinel-2 image was applied to the 2017 satellite data (RMSE = 16.97). The linear models developed for the Sentinel-2 and WorldView-2 images from 2018 were used to map wild leek cover for Gatineau Park. Both images allowed production of similar wild leek maps that, based on field experience and visual inspection of the imagery, provide good descriptions of the actual distribution of wild leek at Gatineau Park.

Recentemente, o uso dos Veículos Aéreos Não Tripulados (VANTs) surgiu como uma ferramenta para a aquisição de dados geoespaciais. Os Modelos Digitais de Elevação (MDEs), as ortoimagens e os modelos tridimensionais gerados a partir de imagens de VANT são produtos cartográficos de grande utilidade para as mais diversas aplicações. Em vista das limitações dos VANTs e da recentidade desta ferramenta, este estudo visa determinar a correlação da configuração dos pontos de apoio e dos parâmetros do voo com a acurácia de ortofotomosaicos elaborados a partir de dados de VANT. Esta pesquisa utilizou dados de três levantamentos aerofotogramétricos distintos realizados com três aeronaves diferentes e considerou parâmetros relacionados à configuração dos pontos de apoio e à execução do voo. Foram gerados 200 ortofotomosaicos e 200 MDEs. Ao todo foram feitas 4616 observações de pontos de controle. Os resultados mostraram baixa correlação linear entre acurácia planimétrica e altimétrica. Os melhores resultados foram obtidos de forma inversa às alturas de voo. No geral, sob as condições de processamento utilizadas neste trabalho, recomenda-se o uso de 4 a 5 pontos de apoio por km², além do uso de voo cruzado. / Recently, the use of Unmanned Aerial Vehicle (UAV) emerged as a tool for geospatial data acquisition. Digital Elevation Models (DEMs), orthoimages and three- dimensional models generated from UAV images are cartographic products of great utility for many different applications. Considering the recency and limitations of UAV, this study aims at determining the correlation of ground control points and flight configuration with the accuracy of orthophotomosaics based on UAV data. This research used data from three different aerial surveys, performed with three different aircrafts, and considered parameters related to ground-control points distribution and to flight missions. 200 orthophotomosaics and 200 DEMs were generated and a total of 4616 ground-control points measurements were performed. Results did not show a linear correlation between planimetric and altimetric accuracy. The best correlation results were obtained inversely related to flight height. In general, under the processing conditions used in this work, we recommend the use of 4 control points per km² and a cross flight pattern.

In this research we aim to achieve cybernetic cohesion information flow in precision agriculture, integrating machine learning methods, computer vision, geographical information system and UAV-photogrammetry in an irrigated area with slaughterhouse wastewater, under five treatments (W100 - irrigation with superficial water and 100% of nitrogen mineral fertilization, E0, E33, E66 and E100 - irrigation with treated effluent from slaughterhouse and addition of 0, 33, 66 and 100% of nitrogen mineral fertilization, respectively) and four replications on grassland (Cynodon dactylon (L.) Pers.). Several images (between one hundred and two hundred) with red, green, blue (RGB) color model were captured using a quadcopter flying at 20 meter altitude and obtaining spatial resolution of 1 centimeter on a surface of approximately 0.5 ha. The images were orthorectified together with nine ground control points done by differential global positioning system (GPS), both processed in the Agisoft PhotoScan software. Thirteen photogrammetric projects were done over time with 30-day revisit, the root mean squared error (RMSE) was used as accuracy measurement, and reached values lower than 5 centimeters for x, y and z axis. The orthoimage obtained with unmanned aerial vehicle (UAV) photogrammetry was changed from RGB to hue, saturation, value (HSV) color model, and the hue color space was chosen due to independence of illumination, beyond it has a good description of exposure of soil and vegetation, but it is dependent of light source temperature, so difficult to estabilish a static threshold, so we selected an unsupervised classification method, K-Means, to classify the unknown patterns along the area. Polygons were drawn delimiting the area represented by each portion and a supervised classification method based on entropy was used, the decision tree, to explore and find patterns that recognize each treatment. These steps are also displayed in forms of georeferenced thematic maps and were executed in the open source softwares Python, QGIS and Weka. The rules defined on the hue color space reached an accuracy of 100% on the training set, and provided a better understanding about the distribution of soil and vegetation on the parcels. This methodology shows a great potential for analysis of spectral data in precision agriculture. / Nesta pesquisa pretendemos alcançar a coesão cibernética no fluxo de informações dentro da agricultura de precisão, integrando métodos de aprendizagem de máquinas, visão computacional, sistema de informação geográfica e aerofotogrametria em uma área irrigada com efluente de matadouro, sob cinco tratamentos (W100 - irrigação com água superficial e 100 % de adubação mineral nitrogenada, E0, E33, E66 e E100 - irrigação com efluente tratado de abatedouro e adição de 0, 33, 66 e 100% de adubação mineral nitrogenada, respectivamente) e quatro repetições em pastagem (Cynodon dactylon (L.) Pers.) Várias imagens (entre cem e duzentas) com modelo de cor vermelho, verde e azul (RGB) foram capturadas por um quadricóptero voando a 20 metros de altitude, e obtendo resolução espacial de 1 centímetro em uma superfície de aproximadamente 0.5 ha. As imagens foram ortorretificadas juntamente com nove pontos de controle, realizados pelo sistema de posicionamento global diferencial (GPS), ambos processados no software Agisoft PhotoScan. Treze projetos fotogramétricos foram realizados ao longo do tempo com revisita de 30 dias, a raiz do erro quadrático médio (RMSE) foi usada como medida de acurácia e atingiu valores menores que 5 centímetros para os eixos x, y e z. A ortoimagem obtida com a fotogrametria do veículo aéreo não tripulado (UAV) foi alterada de RGB para matiz, saturação, valor (HSV) e o espaço de cor matiz foi escolhido devido a independência da iluminação, além de ter boa descrição da exposição do solo e vegetação. Entretanto este é dependente da temperatura da fonte de luz, portanto difícil de se estabelecer um limiar estático, logo selecionamos um método de classificação não supervisionado, o K-Means, para classificar os padrões desconhecidos ao longo da área. Polígonos foram traçados delimitando a área representada por cada parcela e um método supervisionado de classificação baseado na entropia foi utilizado, a árvore de decisão, para explorar e encontrar padrões que reconheçam cada tratamento. Essas etapas também são exibidas em formas de mapas temáticos georeferenciados e foram executadas nos softwares de código aberto Python, QGIS e Weka. As regras definidas no espaço de cor matiz atingiram uma acurácia de 100% no conjunto de treinamento e proporcionaram um melhor entendimento sobre a distribuição do solo e da vegetação nas parcelas. Esta metodologia mostra um grande potencial para análise de dados na agricultura de precisão.

In this research we aim to achieve cybernetic cohesion information flow in precision agriculture, integrating machine learning methods, computer vision, geographical information system and UAV-photogrammetry in an irrigated area with slaughterhouse wastewater, under five treatments (W100 - irrigation with superficial water and 100% of nitrogen mineral fertilization, E0, E33, E66 and E100 - irrigation with treated effluent from slaughterhouse and addition of 0, 33, 66 and 100% of nitrogen mineral fertilization, respectively) and four replications on grassland (Cynodon dactylon (L.) Pers.). Several images (between one hundred and two hundred) with red, green, blue (RGB) color model were captured using a quadcopter flying at 20 meter altitude and obtaining spatial resolution of 1 centimeter on a surface of approximately 0.5 ha. The images were orthorectified together with nine ground control points done by differential global positioning system (GPS), both processed in the Agisoft PhotoScan software. Thirteen photogrammetric projects were done over time with 30-day revisit, the root mean squared error (RMSE) was used as accuracy measurement, and reached values lower than 5 centimeters for x, y and z axis. The orthoimage obtained with unmanned aerial vehicle (UAV) photogrammetry was changed from RGB to hue, saturation, value (HSV) color model, and the hue color space was chosen due to independence of illumination, beyond it has a good description of exposure of soil and vegetation, but it is dependent of light source temperature, so difficult to estabilish a static threshold, so we selected an unsupervised classification method, K-Means, to classify the unknown patterns along the area. Polygons were drawn delimiting the area represented by each portion and a supervised classification method based on entropy was used, the decision tree, to explore and find patterns that recognize each treatment. These steps are also displayed in forms of georeferenced thematic maps and were executed in the open source softwares Python, QGIS and Weka. The rules defined on the hue color space reached an accuracy of 100% on the training set, and provided a better understanding about the distribution of soil and vegetation on the parcels. This methodology shows a great potential for analysis of spectral data in precision agriculture. / Nesta pesquisa pretendemos alcançar a coesão cibernética no fluxo de informações dentro da agricultura de precisão, integrando métodos de aprendizagem de máquinas, visão computacional, sistema de informação geográfica e aerofotogrametria em uma área irrigada com efluente de matadouro, sob cinco tratamentos (W100 - irrigação com água superficial e 100 % de adubação mineral nitrogenada, E0, E33, E66 e E100 - irrigação com efluente tratado de abatedouro e adição de 0, 33, 66 e 100% de adubação mineral nitrogenada, respectivamente) e quatro repetições em pastagem (Cynodon dactylon (L.) Pers.) Várias imagens (entre cem e duzentas) com modelo de cor vermelho, verde e azul (RGB) foram capturadas por um quadricóptero voando a 20 metros de altitude, e obtendo resolução espacial de 1 centímetro em uma superfície de aproximadamente 0.5 ha. As imagens foram ortorretificadas juntamente com nove pontos de controle, realizados pelo sistema de posicionamento global diferencial (GPS), ambos processados no software Agisoft PhotoScan. Treze projetos fotogramétricos foram realizados ao longo do tempo com revisita de 30 dias, a raiz do erro quadrático médio (RMSE) foi usada como medida de acurácia e atingiu valores menores que 5 centímetros para os eixos x, y e z. A ortoimagem obtida com a fotogrametria do veículo aéreo não tripulado (UAV) foi alterada de RGB para matiz, saturação, valor (HSV) e o espaço de cor matiz foi escolhido devido a independência da iluminação, além de ter boa descrição da exposição do solo e vegetação. Entretanto este é dependente da temperatura da fonte de luz, portanto difícil de se estabelecer um limiar estático, logo selecionamos um método de classificação não supervisionado, o K-Means, para classificar os padrões desconhecidos ao longo da área. Polígonos foram traçados delimitando a área representada por cada parcela e um método supervisionado de classificação baseado na entropia foi utilizado, a árvore de decisão, para explorar e encontrar padrões que reconheçam cada tratamento. Essas etapas também são exibidas em formas de mapas temáticos georeferenciados e foram executadas nos softwares de código aberto Python, QGIS e Weka. As regras definidas no espaço de cor matiz atingiram uma acurácia de 100% no conjunto de treinamento e proporcionaram um melhor entendimento sobre a distribuição do solo e da vegetação nas parcelas. Esta metodologia mostra um grande potencial para análise de dados na agricultura de precisão.

As small unmanned aerial systems (UAS) are introduced into the national airspace, measures must be introduced to ensure that they do not interfere with manned aviation and other UAS. Radar provides an attractive solution because of its inherent range accuracy and because it works in diverse weather and lighting conditions. Traditional radar systems, however, are large and high power and do not meet the size, weight and power (SWaP) constraints imposed by UAS, and fully integrated automotive solution do not provide the necessary range. This thesis proposes a compact radar system that meets both the SWaP and range requirements for UAS and can act as a standalone sensor for a sense and avoid system (SAA). The system meets the field of view requirements motivated by the UAS sensing problem (120deg x 30deg) and tracks targets in range and azimuthal angle using a four element phased array receiver. The phased array receiver implements real time correlation and beamforming using a field programmable gate array (FPGA) and can track multiple targets simultaneously. Excluding antennas, the radar transceiver and signal processing platform weighs approximately 120g and is approximately the size of a whiteboard eraser (2.25in x 4in x 1in), which meets the payload requirements of many small (<25kg) UAS. To our knowledge, this is the first real time phased array radar that meets the sensing and SWaP requirements for small UAS.Our testing was done with the radar system on the ground, aimed at airborne UAS targets. Using antennas with a gain of 12 dB, and 800 milliwatts of transmitted power, the system detects UAS targets with a radar cross section of less than 0.1 square meters up to 150 meters away. The ground based system demonstrates radar detectability of extremely small UAS targets, and is scalable to further ranges by increasing antenna gain or adding additional elements. Based on our success in detecting airborne UAS, we conclude that radar remains a feasible option for a UAS collision avoidance sensor.

This thesis presents some existing gimbal and UAV control algorithms as well as novel algorithms developed as the extensions of the existing ones. The existing image-based visual servoing algorithms for both gimbal and UAV require the depth information to the object of interest. The depth information is not measurable when only a monocular camera is used for tracking. This thesis is the result of contemplation to the question: how can the necessity for a depth measurement be removed? A novel gimbal algorithm using adaptive control is developed and presented with simulation and hardware results. Although the estimated depth using the algorithm cannot be used as reliable depth information, the target tracking objective is met. Also, a new UAV control algorithm for target following is developed and presented with simulation results. This algorithm does not require the depth to the target or the UAV altitude to be measured because it exploits the unit vectors to the target and to the optical axis.

Tese de mestrado. Mestrado Integrado em Engenharia Electrotécnica e de Computadores. Faculdade de Engenharia. Universidade do Porto. 2012