| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 5.26 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
Na agricultura de precisão é necessária uma plataforma de aquisição de dados ágil que pode navegar terreno vasto e complexo.
Os Quad-rotores têm grande capacidade de manobra mas geralmente têm autonomia curta, limitando a sua utilidade em vários cenários.
Como a autonomia é um problema de armazenamento de energia e as soluções são normalmente evolutivas, ao invés de revolucionárias, um sistema que pode restaurar a energia armazenada de forma autónoma é uma melhoria na operação de um avião não-tripulado autónoma.
Como o trabalho atual é de natureza concetual, uma configuração de simulação foi desenvolvida para fazer validação básica. O objetivo deste trabalho é integrar tecnologias já existentes. Assim um aspecto importante, é a interface entre componentes do simulador que nunca foram projetados para trabalhar juntos.
Exemplos dessa integração são o processamento de gráficos de uma cena 3D com os dados do motor de física e a exportação de modelos 3D do programa CAD Solidworks. Então, um fluxo de trabalho foi concebido para agregar diferentes tipos de dados. O fluxo de trabalho também foi desenvolvido para criar um ambiente no qual cada componente é o mais dissociado possível de outros outros. Todos os dados simulados alimentados ao sensor vêm do simulador de cinemática, exceto para a câmera. Para o caso da câmera, os dados precisam ser processados para um ambiente sintético 3D. Exceto também para a câmera, os dados são usados pelo sistema de controle de dinâmica de vôo (FDCS), especificamente Arducopter. O módulo Arducopter executa voos no ambiente sintético como faria num real.
Um módulo adicional é ligado ao FDCS que aciona um plano de voo especial para aterrar visualmente em posições predeterminadas. Estes locais têm uma estação que permitem seguir pistas visuais e a troca de bateria. É mostrado que um controlador PI modificado é suficiente fazer uma aterragem visual. Também é mostrado que as tecnologias off-the-shelf podem ser usadas em conjunto com outras especialmente desenvolvidas para criar plataformas de teste estaveis, permitindo a substituição seletiva de funcionalidade. Desta forma o trabalho pôde focar-se nos problemas a serem resolvidos de forma mais abstrata, em vez de implementar recursos que são detalhes práticos. A biblioteca de software MAVLink desenvolvida foi especialmente útil porque criou a ponte para integrar o planeamento de vôo especial desenvolvido e, controlar um módulo de piloto automático de quad-rotor em vôo.
In precision agriculture there is a need for an agile data acquisition platform that can navigate vast and complex terrain. Small sized electrical rotor-craft have great maneuverability but generally have short endurance, limiting their roles in several scenarios. As endurance is a power storage problem, and solutions are normally evolutionary rather than revolutionary, a system that can restore it's relatively low power capacity while remaining autonomous is an improvement in the operation of an autonomous unmanned airplane. As the current work is of a conceptual nature, a simulation setup was developed to do basic validation. The goal of this work is to integrate already existing technologies thus, an important aspect, is the interface between simulator components that were never designed to work together. Examples of this integration are the graphics rendering of a 3D scene with data from the physics engine and the exporting of 3D models from Solidworks CAD program. A workflow was then devised to aggregate different types of data. The workflow was also developed to create an environment in which each component is the most decoupled from each other. All the simulated data fed to the sensor models come from the kinematics simulator except for the camera. For the case of the camera, data needs to be rendered to a 3D scene. Except for the camera, data is used by the Flight Dynamics Control System module (FDCS), specifically Arducopter. The Arducopter module executes flights in the synthetic environment as it would in a real one. An additional module is connected to the FDCS that triggers a special flight plan to land visually in predetermined locations. These locations have a station that allows for the tracking of visual cues and battery switching. It is shown that a modified PI controller is enough make a visual landing. It is also shown that off-the shelf technologies can be used in conjunction with custom developed ones to create stable testing platforms, allowing for selective replacement of functionality. This way work can focus on the problems to be solved in a more abstract way instead of having to implement features which are practical details. The developed MAVLink library was specially useful because it created the bridge to integrate the developed landing flight planning and control into a general quad-rotor autopilot module in flight.
In precision agriculture there is a need for an agile data acquisition platform that can navigate vast and complex terrain. Small sized electrical rotor-craft have great maneuverability but generally have short endurance, limiting their roles in several scenarios. As endurance is a power storage problem, and solutions are normally evolutionary rather than revolutionary, a system that can restore it's relatively low power capacity while remaining autonomous is an improvement in the operation of an autonomous unmanned airplane. As the current work is of a conceptual nature, a simulation setup was developed to do basic validation. The goal of this work is to integrate already existing technologies thus, an important aspect, is the interface between simulator components that were never designed to work together. Examples of this integration are the graphics rendering of a 3D scene with data from the physics engine and the exporting of 3D models from Solidworks CAD program. A workflow was then devised to aggregate different types of data. The workflow was also developed to create an environment in which each component is the most decoupled from each other. All the simulated data fed to the sensor models come from the kinematics simulator except for the camera. For the case of the camera, data needs to be rendered to a 3D scene. Except for the camera, data is used by the Flight Dynamics Control System module (FDCS), specifically Arducopter. The Arducopter module executes flights in the synthetic environment as it would in a real one. An additional module is connected to the FDCS that triggers a special flight plan to land visually in predetermined locations. These locations have a station that allows for the tracking of visual cues and battery switching. It is shown that a modified PI controller is enough make a visual landing. It is also shown that off-the shelf technologies can be used in conjunction with custom developed ones to create stable testing platforms, allowing for selective replacement of functionality. This way work can focus on the problems to be solved in a more abstract way instead of having to implement features which are practical details. The developed MAVLink library was specially useful because it created the bridge to integrate the developed landing flight planning and control into a general quad-rotor autopilot module in flight.
Description
Keywords
Aproximação Visual. Arducopter Atracagem Quad-Rotor Troca de Bateria Voo Autonomo