Aktionenlernen mit Selbstorganisierenden Karten und Reinforcement Learning

Die vorliegende Arbeit beschäftigt sich mit der Entwicklung eines Funktionsapproximators und dessen Verwendung in Verfahren zum Lernen von diskreten und kontinuierlichen Aktionen: 1. Ein allgemeiner Funktionsapproximator – Locally Weighted Interpolating Growing Neural Gas (LWIGNG) – wird auf Basis eines Wachsenden Neuralen Gases (GNG) entwickelt. Die topologische Nachbarschaft in der Neuronenstruktur wird verwendet, um zwischen benachbarten Neuronen zu interpolieren und durch lokale Gewichtung die Approximation zu berechnen. Die Leistungsfähigkeit des Ansatzes, insbesondere in Hinsicht auf sich verändernde Zielfunktionen und sich verändernde Eingabeverteilungen, wird in verschiedenen Experimenten unter Beweis gestellt. 2. Zum Lernen diskreter Aktionen wird das LWIGNG-Verfahren mit Q-Learning zur Q-LWIGNG-Methode verbunden. Dafür muss der zugrunde liegende GNG-Algorithmus abgeändert werden, da die Eingabedaten beim Aktionenlernen eine bestimmte Reihenfolge haben. Q-LWIGNG erzielt sehr gute Ergebnisse beim Stabbalance- und beim Mountain-Car-Problem und gute Ergebnisse beim Acrobot-Problem. 3. Zum Lernen kontinuierlicher Aktionen wird ein REINFORCE-Algorithmus mit LWIGNG zur ReinforceGNG-Methode verbunden. Dabei wird eine Actor-Critic-Architektur eingesetzt, um aus zeitverzögerten Belohnungen zu lernen. LWIGNG approximiert sowohl die Zustands-Wertefunktion als auch die Politik, die in Form von situationsabhängigen Parametern einer Normalverteilung repräsentiert wird. ReinforceGNG wird erfolgreich zum Lernen von Bewegungen für einen simulierten 2-rädrigen Roboter eingesetzt, der einen rollenden Ball unter bestimmten Bedingungen abfangen soll.

Analysis and 2D Simulation of a Hexapod Robot Leg for Remote Exploration

A walking machine is a wheeled rover alternative, well suited for work in an unstructured environment and specially in abrupt terrain. They have some drawback like speed and power consumption, but they can achieve complex movements and protrude very little the environment they are working on. The locomotion system is determined by the terrain conditions and, in our case, this legged design has been chosen based in a working area like Rio Tinto in the South of Spain, which is a river area with abrupt terrain. A walking robot with so many degrees of freedom can be a challenge when dealing with the analysis and simulations of the legs. This paper shows how to deal with the kinematical analysis of the equations of a hexapod robot based on a design developed by the Center of Astrobiology INTA-CSIC following the classical formulation of equations

User-independent accelerometer-based gesture recognition for mobile devices

Many mobile devices embed nowadays inertial sensors. This enables new forms of human-computer interaction through the use of gestures (movements performed with the mobile device) as a way of communication. This paper presents an accelerometer-based gesture recognition system for mobile devices which is able to recognize a collection of 10 different hand gestures. The system was conceived to be light and to operate in a user -independent manner in real time. The recognition system was implemented in a smart phone and evaluated through a collection of user tests, which showed a recognition accuracy similar to other state-of-the art techniques and a lower computational complexity. The system was also used to build a human -robot interface that enables controlling a wheeled robot with the gestures made with the mobile phone.

ARGoS: a modular, multi-engine simulator for heterogeneous swarm robotics

We present ARGoS, a novel open source multi-robot simulator. The main design focus of ARGoS is the real-time simulation of large heterogeneous swarms of robots. Existing robot simulators obtain scalability by imposing limitations on their extensibility and on the accuracy of the robot models. By contrast, in ARGoS we pursue a deeply modular approach that allows the user both to easily add custom features and to allocate computational resources where needed by the experiment. A unique feature of ARGoS is the possibility to use multiple physics engines of different types and to assign them to different parts of the environment. Robots can migrate from one engine to another transparently. This feature enables entirely novel classes of optimizations to improve scalability and paves the way for a new approach to parallelism in robotics simulation. Results show that ARGoS can simulate about 10,000 simple wheeled robots 40% faster than real-time.

Analysis and optimization for a hexapod walking robot for planetary missions

La tecnología de las máquinas móviles autónomas ha sido objeto de una gran investigación y desarrollo en las últimas décadas. En muchas actividades y entornos, los robots pueden realizar operaciones que son duras, peligrosas o simplemente imposibles para los humanos. La exploración planetaria es un buen ejemplo de un entorno donde los robots son necesarios para realizar las tareas requeridas por los científicos. La reciente exploración de Marte con robots autónomos nos ha mostrado la capacidad de las nuevas tecnologías. Desde la invención de la rueda, que esta acertadamente considerado como el mayor invento en la historia del transporte humano, casi todos los vehículos para exploración planetaria han empleado las ruedas para su desplazamiento. Las nuevas misiones planetarias demandan maquinas cada vez mas complejas. En esta Tesis se propone un nuevo diseño de un robot con patas o maquina andante que ofrecerá claras ventajas en entornos extremos. Se demostrara que puede desplazarse en los terrenos donde los robots con ruedas son ineficientes, convirtiéndolo en una elección perfecta para misiones planetarias. Se presenta una reseña histórica de los principales misiones espaciales, en particular aquellos dirigidos a la exploración planetaria. A través de este estudio será posible analizar las desventajas de los robots con ruedas utilizados en misiones anteriores. El diseño propuesto de robot con patas será presentado como una alternativa para aquellas misiones donde los robots con ruedas puedan no ser la mejor opción. En esta tesis se presenta el diseño mecánico de un robot de seis patas capaz de soportar las grandes fuerzas y momentos derivadas del movimiento de avance. Una vez concluido el diseño mecánico es necesario realizar un análisis que permita entender el movimiento y comportamiento de una maquina de esta complejidad. Las ecuaciones de movimiento del robot serán validadas por dos métodos: cinemático y dinámico. Dos códigos Matlab® han sido desarrollados para resolver dichos sistemas de ecuaciones y han sido verificados por un tercer método, un modelo de elementos finitos, que también verifica el diseño mecánico. El robot con patas presentado, ha sido diseñado para la exploración planetaria en Marte. El comportamiento del robot durante sus desplazamientos será probado mediante un código de Matlab®, desarrollado para esta tesis, que permite modificar las trayectorias, el tipo de terreno, y el número y altura de los obstáculos. Estos terrenos y requisitos iniciales no han sido elegidos de forma aleatoria, si no que están basados en mi experiencia como miembro del equipo de MSL-NASA que opera un instrumento a bordo del rover Curiosity en Marte. El robot con patas desarrollado y fabricado por el Centro de Astrobiología (INTA-CSIC), esta basado en el diseño mecánico y análisis presentados en esta tesis. ABSTRACT The autonomous machines technology has undergone a major research and development during the last decades. In many activities and environments, robots can perform operations that are tought, dangerous or simply imposible to humans. Planetary exploration is a good example of such environment where robots are needed to perform the tasks required by the scientits. Recent Mars exploration based on autonomous vehicles has shown us the capacity of the new technologies. From the invention of the wheel, which is rightly regarded as the greatest invention in the history of human transportation, nearly all-planetary vehicles are based in wheeled locomotion, but new missions demand new types of machines due to the complex tasks needed to be performed. It will be proposed in this thesis a new design of a legged robot or walking machine, which may offer clear advantages in tough environments. This Thesis will show that the proposed walking machine can travel, were terrain difficulties make wheeled vehicles ineffective, making it a perfect choice for planetary mission. A historical background of the main space missions, in particular those aimed at planetary exploration will be presented. From this study the disadvantages found in the existing wheel rovers will be analysed. The legged robot designed will be introduced as an alternative were wheeled rovers could be no longer the best option for planetary exploration. This thesis introduces the mechanical design of a six-leg robot capable of withstanding high forces and moments due to the walking motion. Once the mechanical design is concluded, and in order to analyse a machine of this complexity an understanding of its movement and behaviour is mandatory. This movement equation will be validated by two methods: kinematics and dynamics. Two Matlab® codes have been developed to solve the systems of equations and validated by a third method, a finite element model, which also verifies the mechanical design. The legged robot presented has been designed for a Mars planetary exploration. The movement behaviour of the robot will be tested in a Matlab® code developed that allows to modify the trajectories, the type of terrain, number and height of obstacles. These terrains and initial requirements have not been chosen randomly, those are based on my experience as a member of the MSL NASA team, which operates an instrument on-board of the Curiosity rover in Mars. The walking robot developed and manufactured by the Center of Astrobiology (CAB) is based in the mechanical design and analysis that will be presented in this thesis.

Injury risk from popular childhood physical activities: results from an Australian primary school cohort

Background: Children engage in various physical activities that pose different injury risks. However, the lack of adequate data on exposure has meant that these risks have not been quantified or compared in young children aged 5-12 years. Objectives: To measure exposure to popular activities among Australian primary school children and to quantify the associated injury risks. Method: The Childhood Injury Prevention Study prospectively followed up a cohort of randomly selected Australian primary and preschool children aged 5-12 years. Time (min) engaged in various physical activities was measured using a parent-completed 7-day diary. All injuries over 12 months were reported to the study. All data on exposure and injuries were coded using the International classification of external causes of injury. Injury rates per 1000 h of exposure were calculated for the most popular activities. Results: Complete diaries and data on injuries were available for 744 children. Over 12 months, 314 injuries relating to physical activity outside of school were reported. The highest injury risks per exposure time occurred for tackle-style football (2.18/1000 h), wheeled activities (1.72/1000 h) and tennis (1.19/1000 h). Overall, boys were injured more often than girls; however, the differences were non-significant or reversed for some activities including soccer, trampolining and team ball sports. Conclusion: Although the overall injury rate was low in this prospective cohort, the safety of some popular childhood activities can be improved so that the benefits may be enjoyed with fewer negative consequences.

Vulnerability and disobedience:new repertoires in French environmental protests

The emergence of the counter-globalisation movement in France has been accompanied by an apparent diversification of social protest repertoires. Protest events carried out by groups associated with a wide array of issues have been remarkable for their use of spectacular and novel actions, while civil disobedience campaigns have been prominent features of environmental and civil rights protests in particular. Drawing on a number of examples of contemporary environmental and global justice campaigns, opposing advertising, four-wheeled drive vehicles, nuclear energy and, especially, open field trials of genetically modified crops, this article discusses the rise of such new forms of protest, placing them in the wider context of transformations in protest repertoires in France. It identifies key examples of innovation, before discussing the twin processes of diffusion and domestication that shape them. It is argued that, although transnational agents and processes are key determinants of repertoire innovation, it is vital to identify the national, movement and sectoral contexts and discourses which enable the naturalisation and legitimisation of new action forms.

Pose estimation of a humanoid robot using images from a mobile external camera

NOGUEIRA, Marcelo B. ; MEDEIROS, Adelardo A. D. ; ALSINA, Pablo J. Pose Estimation of a Humanoid Robot Using Images from an Mobile Extern Camera. In: IFAC WORKSHOP ON MULTIVEHICLE SYSTEMS, 2006, Salvador, BA. Anais... Salvador: MVS 2006, 2006.

Pose estimation of a humanoid robot using images from a mobile external camera

NOGUEIRA, Marcelo B. ; MEDEIROS, Adelardo A. D. ; ALSINA, Pablo J. Pose Estimation of a Humanoid Robot Using Images from an Mobile Extern Camera. In: IFAC WORKSHOP ON MULTIVEHICLE SYSTEMS, 2006, Salvador, BA. Anais... Salvador: MVS 2006, 2006.

Multirobot Tethering for Localization and Control

Particle filtering has proven to be an effective localization method for wheeled autonomous vehicles. For a given map, a sensor model, and observations, occasions arise where the vehicle could equally likely be in many locations of the map. Because particle filtering algorithms may generate low confidence pose estimates under these conditions, more robust localization strategies are required to produce reliable pose estimates. This becomes more critical if the state estimate is an integral part of system control. We investigate the use of particle filter estimation techniques on a hovercraft vehicle. The marginally stable dynamics of a hovercraft require reliable state estimates for proper stability and control. We use the Monte Carlo localization method, which implements a particle filter in a recursive state estimate algorithm. An H-infinity controller, designed to accommodate the latency inherent in our state estimation, provides stability and controllability to the hovercraft. In order to eliminate the low confidence estimates produced in certain environments, a multirobot system is designed to introduce mobile environment features. By tracking and controlling the secondary robot, we can position the mobile feature throughout the environment to ensure a high confidence estimate, thus maintaining stability in the system. A laser rangefinder is the sensor the hovercraft uses to track the secondary robot, observe the environment, and facilitate successful localization and stability in motion.

Effect of different methods of soil preparation on soil moisture loss, weed growth and soil resistance to penetration.

The effect of different methods of soil preparation on soil moisture loss, weed growth and soil resistance to penetration has been quantified for the Latosols of Bebedouro experiment station of CPATSA-EMBRAPA, Petrolina,PE, Brazil. The methods studied are manual preparation, preparation by animal drawn wheeled tool carrier and tractorized implements. The drying characteristics of three fields with different soil physical properties were studied prior to conducting the experiment. The different methods of soil preparation caused the soil moisture loss differently. The plot prepared by animal drawn tool carrier retained more moisture for longer time even at deeper layers. The soil resistance to penetration on the surface increased with time with little variation at deeper layers.

Lamina-enleiradora a tracao animal e seu uso em novo sistema de captacao de agua de chuva in situ.

A lamina-enleiradora a tracao animal e um equipamento para uso como conexo de chassis-porta-implementos, adaptado a um sulcador tipo asas extensas. Para se adaptar este sulcador como lamina-enleiradora, precisa-se trocar uma das asas por uma chapa de ferro de 30 cm x 110 cm. A capacidade operacional deste equipamento varia de 0,16 a 0,47 ha/h, durante instalacao, reformacao ou capina num sistema inovador de captacao de agua de chuva in situ. Este sistema consiste de camalhoes largos e estreitos alternados em curva de nivel. Os camalhoes largos de forma triangular, servem como area de captacao; os estreitos, como area de plantio, e o espaco entre os dois, como area de armazenamento. A relacao entre area de captacao e de plantio e de 2:1.

O conceito de armas combinadas e a compatibilidade dos equipamentos do Exército Português

presente Trabalho de Investigação Aplicada, incide sobre a temática do Agrupamento de Armas Combinadas “LEODUR”, um Agrupamento que emprega o Carro de Combate Leopard 2A6, com a Viatura Blindada de Rodas Pandur II (8x8) Infantry Carrier Vehicle e a Viatura Blindada de Rodas Pandur II (8x8) Infantry Fighting Vehicle, num contexto doutrinário de Armas Combinadas. Devido ao facto de algumas destas viaturas serem de lagartas e outras serem de rodas, a sua combinação poderá não trazer maiores vantagens do que aquelas que cada meio traz se atuar separadamente. É nesta problemática que se pretende investigar quais as potencialidades e vulnerabilidades do uso simultâneo destes meios e se valerá realmente a pena utiliza-los de forma combinada. Pretende-se com esta Investigação, explicar de que forma as viaturas Pandur II (8x8) e o Leopard 2A6 são compatíveis num contexto de Armas Combinadas em Território Nacional, numa unidade de escalão Batalhão (Agrupamento). A investigação em si segue as orientações do Método Hipotético-Dedutivo, pelo que serão levantadas um conjunto de hipóteses a ser confirmadas e/ou infirmadas. A recolha de dados será principalmente através de documentos, como manuais doutrinários, revistas, inquéritos por entrevistas e ainda pela observação direta do Exercício Lince 2016. A aplicação simultânea destas duas tipologias de viaturas traduz-se num enorme Poder de Fogo, elevada Proteção e Mobilidade para um Agrupamento, com a valência de utilizarem meios de Comando e Controlo iguais, possibilitando maior rapidez nas comunicações e uma permanente visualização do Campo de Batalha.

Adequabilidade das VBR PANDUR II 8x8

O intento predominante desta investigação prende-se em tomar conhecimento de como as recentes Viaturas Blindadas de Rodas adquiridas pelo Exército Português se adequam em Território Nacional, naquilo que se poderá constituir como um Eixo de Aproximação, e inseridas numa operação defensiva. Para a realização deste trabalho foi adotada uma metodologia de investigação baseada em alguns autores. Para a recolha de informação foi usada a pesquisa documental e bibliográfica e a realização de entrevistas. Este trabalho encontra-se dividido em várias partes, sendo que se dividem em dois grandes grupos. Numa primeira fase são introduzidos conceitos teóricos que suportam a investigação, nomeadamente no âmbito da Zona de Operações Terrestre, e das características das diferentes versões da VBR PANDUR II 8x8. Numa segunda fase, conduz-se o Estudo do Espaço de Batalha pelas Informações numa determinada região do Território Nacional, terminando esta etapa com a junção dos conhecimentos anteriormente adquiridos com a análise do terreno feita. Concluiu-se com a realização desta investigação que as forças equipadas com estas viaturas estão preparadas para executar operações de índole defensiva no âmbito das operações retrógradas ou defesas móveis, devido às suas capacidades. Com a sua elevada capacidade de projeção, mobilidade do terreno, autonomia, observação do campo de batalha, estas viaturas conferem ao Exército Português uma nova tipologia de força – A Força Média.

Estudo sobre as principais características das viaturas táticas ligeiras blindadas para o Exército português.

O presente Trabalho de Investigação Aplicada tem como estudo identificar as principais características das viaturas táticas ligeiras blindadas, que irão equipar o Exército Português, por forma a garantir a mobilidade tática terrestre. Com esta investigação pretende-se identificar a configuração do ambiente operacional, onde as viaturas das Forças Nacionais Destacadas tem sido empregues. De igual modo identificar as caraterísticas das viaturas táticas ligeiras blindadas disponíveis no mercado da indústria de defesa passíveis de serem adquiridas, descrevendo ainda o tipo de operações em que as forças do Exército Português tem participado, utilizando viaturas táticas ligeiras. Esta investigação tem como principal objetivo identificar as principais caraterísticas das viaturas táticas ligeiras blindadas a adquirir pelo Exército Português, para se adaptarem à realidade atual do ambiente operacional onde Portugal participa com forças, e fazer uma caracterização da mais recente viatura, a Pandur II 8x8, e a viatura com mais presenças em teatros de operações, o High Mobility Multipurpose Wheeled Vehicle. Para alcançar este objetivo principal, foi utilizado o método dedutivo partindo do ambiente operacional dos diversos Teatros de Operações em que Portugal tem empregue forças, para as principais caraterísticas das viaturas. O ambiente operacional que reside hoje em dia, nos teatros de operações, é caraterizado por assimetrias, resultantes de novas ameaças. Estas necessitam de esforços, por parte das forças militares para se adaptarem. As caraterísticas das viaturas táticas ligeiras blindadas têm vindo a sofrer algumas alterações na sua proteção e mobilidade, onde apresentam melhorias técnicas, face às inovações tecnológicas, mas há outras, como o poder de fogo que podem vir a ser colmatadas no futuro.