A geração de bons movimentos em tempo real para corpos com muitos graus de liberdade ainda é um desafio. Uma quantidade elevada de graus de liberdade aumenta de forma exponencial a quantidade de posições diferentes que um determinado corpo pode obter. Fazer uso dessa quantidade de possibilidades para gerar movimentos complexos pode ser extremamente útil para planejamento de movimentos de robôs ou personagens virtuais, porém incrivelmente caro em termos computacionais. Existem muitos algoritmos que se baseiam no uso de mapas de caminhos (chamados roadmaps) para trabalhar com corpos com muitos graus de liberdade. Um roadmap funciona como uma coletânea de poses de um corpo interligadas entre si, onde cada ligação representa uma possibilidade de transição livre de colisões. Geralmente as técnicas que utilizam roadmaps usam abordagens determinísticas ou aleatórias para atingir o objetivo. Através de métodos determinísticos é possível explorar de forma mais uniforme o espaço de configuração, garantindo uma melhor cobertura e qualidade do roadmap. Já as abordagens aleatórias, geralmente permitem um melhor desempenho e, principalmente, tornam viáveis a aplicação de uma solução para corpos com muitos graus de liberdade. Neste trabalho é proposto um método determinístico adaptável para a geração de roadmaps (ADRM) que provê uma cobertura adequada do espaço de configuração em um tempo perfeitamente aceitável em comparação a outros métodos. Para obter isso, é feita em primeiro lugar uma classificação de todos os DOFs do modelo e, então, essa classificação é usada como parâmetro para decidir quantas amostras serão geradas de cada DOF. A combinação entre as amostras de todos os DOFs gera a quantidade total de amostras. Para validação do novo método foram executados diversos testes em ambientes distintos. Os testes foram avaliados através da comparação com outras técnicas existentes, em quesitos como tempo de geração e cobertura do espaço de configuração. Os resultados demonstram que o método atinge uma cobertura do espaço de configuração muito boa, em um tempo aceitável. / The creation of good real time movements for bodies with many degrees of freedom (DOF) still remains a challenge. A great amount of DOFs increase, in an exponential way, the quantity of different positions that a body can assume. Making use of that amount of possibilities to generate complex movements can be useful for planning robots’ movements or even to animate virtual characters, however it is extremely expensive in computational terms. There are many algorithms that are based on the use of roadmaps to work with bodies with many degrees of freedom. A roadmap works as a collection of valid body’s positions interconnected, where each connection represents a possibility of a transaction free of collisions. Usually, the techniques which make use of roadmaps follow deterministic or probabilistic approaches to get to the objective. Trough deterministic methods it is possible to explore in a more uniform way the configuration’s space, assuring a better covering and quality of the roadmap. Therefore, probabilistic (or random) approaches allow a better performance and, mainly, make possibly the application of a solution for bodies with higher degrees of freedom. This work proposes a deterministic method applicable to roadmaps generation (ADRM) which provides an adequate covering of the configuration’s space in a completely acceptable time range comparing to other rates. To achieve this goal, first of all a classification of all of the DOFs of the model is made and, then, this classification is used as a parameter to decide how many samples will be generated of each DOF. The combining between the samples of all of the DOFs generates the total amount of samples. To validate the new method, several tests were executed at different environments. The tests were evaluated trough the comparison with other existents techniques, using criteria like the time spent in generating a roadmap and covering of the space of configuration. The results show us that the method achieves a satisfactory covering of the space configuration in an acceptable time range.

A geração de bons movimentos em tempo real para corpos com muitos graus de liberdade ainda é um desafio. Uma quantidade elevada de graus de liberdade aumenta de forma exponencial a quantidade de posições diferentes que um determinado corpo pode obter. Fazer uso dessa quantidade de possibilidades para gerar movimentos complexos pode ser extremamente útil para planejamento de movimentos de robôs ou personagens virtuais, porém incrivelmente caro em termos computacionais. Existem muitos algoritmos que se baseiam no uso de mapas de caminhos (chamados roadmaps) para trabalhar com corpos com muitos graus de liberdade. Um roadmap funciona como uma coletânea de poses de um corpo interligadas entre si, onde cada ligação representa uma possibilidade de transição livre de colisões. Geralmente as técnicas que utilizam roadmaps usam abordagens determinísticas ou aleatórias para atingir o objetivo. Através de métodos determinísticos é possível explorar de forma mais uniforme o espaço de configuração, garantindo uma melhor cobertura e qualidade do roadmap. Já as abordagens aleatórias, geralmente permitem um melhor desempenho e, principalmente, tornam viáveis a aplicação de uma solução para corpos com muitos graus de liberdade. Neste trabalho é proposto um método determinístico adaptável para a geração de roadmaps (ADRM) que provê uma cobertura adequada do espaço de configuração em um tempo perfeitamente aceitável em comparação a outros métodos. Para obter isso, é feita em primeiro lugar uma classificação de todos os DOFs do modelo e, então, essa classificação é usada como parâmetro para decidir quantas amostras serão geradas de cada DOF. A combinação entre as amostras de todos os DOFs gera a quantidade total de amostras. Para validação do novo método foram executados diversos testes em ambientes distintos. Os testes foram avaliados através da comparação com outras técnicas existentes, em quesitos como tempo de geração e cobertura do espaço de configuração. Os resultados demonstram que o método atinge uma cobertura do espaço de configuração muito boa, em um tempo aceitável. / The creation of good real time movements for bodies with many degrees of freedom (DOF) still remains a challenge. A great amount of DOFs increase, in an exponential way, the quantity of different positions that a body can assume. Making use of that amount of possibilities to generate complex movements can be useful for planning robots’ movements or even to animate virtual characters, however it is extremely expensive in computational terms. There are many algorithms that are based on the use of roadmaps to work with bodies with many degrees of freedom. A roadmap works as a collection of valid body’s positions interconnected, where each connection represents a possibility of a transaction free of collisions. Usually, the techniques which make use of roadmaps follow deterministic or probabilistic approaches to get to the objective. Trough deterministic methods it is possible to explore in a more uniform way the configuration’s space, assuring a better covering and quality of the roadmap. Therefore, probabilistic (or random) approaches allow a better performance and, mainly, make possibly the application of a solution for bodies with higher degrees of freedom. This work proposes a deterministic method applicable to roadmaps generation (ADRM) which provides an adequate covering of the configuration’s space in a completely acceptable time range comparing to other rates. To achieve this goal, first of all a classification of all of the DOFs of the model is made and, then, this classification is used as a parameter to decide how many samples will be generated of each DOF. The combining between the samples of all of the DOFs generates the total amount of samples. To validate the new method, several tests were executed at different environments. The tests were evaluated trough the comparison with other existents techniques, using criteria like the time spent in generating a roadmap and covering of the space of configuration. The results show us that the method achieves a satisfactory covering of the space configuration in an acceptable time range.

A geração de bons movimentos em tempo real para corpos com muitos graus de liberdade ainda é um desafio. Uma quantidade elevada de graus de liberdade aumenta de forma exponencial a quantidade de posições diferentes que um determinado corpo pode obter. Fazer uso dessa quantidade de possibilidades para gerar movimentos complexos pode ser extremamente útil para planejamento de movimentos de robôs ou personagens virtuais, porém incrivelmente caro em termos computacionais. Existem muitos algoritmos que se baseiam no uso de mapas de caminhos (chamados roadmaps) para trabalhar com corpos com muitos graus de liberdade. Um roadmap funciona como uma coletânea de poses de um corpo interligadas entre si, onde cada ligação representa uma possibilidade de transição livre de colisões. Geralmente as técnicas que utilizam roadmaps usam abordagens determinísticas ou aleatórias para atingir o objetivo. Através de métodos determinísticos é possível explorar de forma mais uniforme o espaço de configuração, garantindo uma melhor cobertura e qualidade do roadmap. Já as abordagens aleatórias, geralmente permitem um melhor desempenho e, principalmente, tornam viáveis a aplicação de uma solução para corpos com muitos graus de liberdade. Neste trabalho é proposto um método determinístico adaptável para a geração de roadmaps (ADRM) que provê uma cobertura adequada do espaço de configuração em um tempo perfeitamente aceitável em comparação a outros métodos. Para obter isso, é feita em primeiro lugar uma classificação de todos os DOFs do modelo e, então, essa classificação é usada como parâmetro para decidir quantas amostras serão geradas de cada DOF. A combinação entre as amostras de todos os DOFs gera a quantidade total de amostras. Para validação do novo método foram executados diversos testes em ambientes distintos. Os testes foram avaliados através da comparação com outras técnicas existentes, em quesitos como tempo de geração e cobertura do espaço de configuração. Os resultados demonstram que o método atinge uma cobertura do espaço de configuração muito boa, em um tempo aceitável. / The creation of good real time movements for bodies with many degrees of freedom (DOF) still remains a challenge. A great amount of DOFs increase, in an exponential way, the quantity of different positions that a body can assume. Making use of that amount of possibilities to generate complex movements can be useful for planning robots’ movements or even to animate virtual characters, however it is extremely expensive in computational terms. There are many algorithms that are based on the use of roadmaps to work with bodies with many degrees of freedom. A roadmap works as a collection of valid body’s positions interconnected, where each connection represents a possibility of a transaction free of collisions. Usually, the techniques which make use of roadmaps follow deterministic or probabilistic approaches to get to the objective. Trough deterministic methods it is possible to explore in a more uniform way the configuration’s space, assuring a better covering and quality of the roadmap. Therefore, probabilistic (or random) approaches allow a better performance and, mainly, make possibly the application of a solution for bodies with higher degrees of freedom. This work proposes a deterministic method applicable to roadmaps generation (ADRM) which provides an adequate covering of the configuration’s space in a completely acceptable time range comparing to other rates. To achieve this goal, first of all a classification of all of the DOFs of the model is made and, then, this classification is used as a parameter to decide how many samples will be generated of each DOF. The combining between the samples of all of the DOFs generates the total amount of samples. To validate the new method, several tests were executed at different environments. The tests were evaluated trough the comparison with other existents techniques, using criteria like the time spent in generating a roadmap and covering of the space of configuration. The results show us that the method achieves a satisfactory covering of the space configuration in an acceptable time range.

En biologie structurale, l’étude théorique de structures moléculaires comporte quatre activités principales organisées selon le processus séquentiel suivant : la collecte de données expérimentales/théoriques, la visualisation des structures 3d, la simulation moléculaire, l’analyse et l’interprétation des résultats. Cet enchaînement permet à l’expert d’élaborer de nouvelles hypothèses, de les vérifier de manière expérimentale et de produire de nouvelles données comme point de départ d’un nouveau processus.L’explosion de la quantité de données à manipuler au sein de cette boucle pose désormais deux problèmes. Premièrement, les ressources et le temps relatifs aux tâches de transfert et de conversion de données entre chacune de ces activités augmentent considérablement. Deuxièmement, la complexité des données moléculaires générées par les nouvelles méthodologies expérimentales accroît fortement la difficulté pour correctement percevoir, visualiser et analyser ces données.Les environnements immersifs sont souvent proposés pour aborder le problème de la quantité et de la complexité croissante des phénomènes modélisés, en particulier durant l’activité de visualisation. En effet, la Réalité Virtuelle offre entre autre une perception stéréoscopique de haute qualité utile à une meilleure compréhension de données moléculaires intrinsèquement tridimensionnelles. Elle permet également d’afficher une quantité d’information importante grâce aux grandes surfaces d’affichage, mais aussi de compléter la sensation d’immersion par d’autres canaux sensorimoteurs.Cependant, deux facteurs majeurs freinent l’usage de la Réalité Virtuelle dans le domaine de la biologie structurale. D’une part, même s’il existe une littérature fournie sur la navigation dans les scènes virtuelles réalistes et écologiques, celle-ci est très peu étudiée sur la navigation sur des données scientifiques abstraites. La compréhension de phénomènes 3d complexes est pourtant particulièrement conditionnée par la capacité du sujet à se repérer dans l’espace. Le premier objectif de ce travail de doctorat a donc été de proposer des paradigmes navigation 3d adaptés aux structures moléculaires complexes. D’autre part, le contexte interactif des environnements immersif favorise l’interaction directe avec les objets d’intérêt. Or les activités de collecte et d’analyse des résultats supposent un contexte de travail en "ligne de commande" ou basé sur des scripts spécifiques aux outils d’analyse. Il en résulte que l’usage de la Réalité Virtuelle se limite souvent à l’activité d’exploration et de visualisation des structures moléculaires. C’est pourquoi le second objectif de thèse est de rapprocher ces différentes activités, jusqu’alors réalisées dans des contextes interactifs et applicatifs indépendants, au sein d’un contexte interactif homogène et unique. Outre le fait de minimiser le temps passé dans la gestion des données entre les différents contextes de travail, il s’agit également de présenter de manière conjointe et simultanée les structures moléculaires et leurs analyses et de permettre leur manipulation par des interactions directes.Notre contribution répond à ces objectifs en s’appuyant sur une approche guidée à la fois par le contenu et la tâche. Des paradigmes de navigation ont été conçus en tenant compte du contenu moléculaire, en particulier des propriétés géométriques, et des tâches de l’expert, afin de faciliter le repérage spatial et de rendre plus performante l’activité d’exploration. Par ailleurs, formaliser la nature des données moléculaires, leurs analyses et leurs représentations visuelles, permettent notamment de proposer à la demande et interactivement des analyses adaptées à la nature des données et de créer des liens entre les composants moléculaires et les analyses associées. Ces fonctionnalités passent par la construction d’une représentation sémantique unifiée et performante rendant possible l’intégration de ces activités dans un contexte interactif unique. / In structural biology, the theoretical study of molecular structures has four main activities organized in the following scenario: collection of experimental and theoretical data, visualization of 3D structures, molecular simulation, analysis and interpretation of results. This pipeline allows the expert to develop new hypotheses, to verify them experimentally and to produce new data as a starting point for a new scenario.The explosion in the amount of data to handle in this loop has two problems. Firstly, the resources and time dedicated to the tasks of transfer and conversion of data between each of these four activities increases significantly. Secondly, the complexity of molecular data generated by new experimental methodologies greatly increases the difficulty to properly collect, visualize and analyze the data.Immersive environments are often proposed to address the quantity and the increasing complexity of the modeled phenomena, especially during the viewing activity. Indeed, virtual reality offers a high quality stereoscopic perception, useful for a better understanding of inherently three-dimensional molecular data. It also displays a large amount of information thanks to the large display surfaces, but also to complete the immersive feeling with other sensorimotor channels (3D audio, haptic feedbacks,...).However, two major factors hindering the use of virtual reality in the field of structural biology. On one hand, although there are literature on navigation and environmental realistic virtual scenes, navigating abstract science is still very little studied. The understanding of complex 3D phenomena is however particularly conditioned by the subject’s ability to identify themselves in a complex 3D phenomenon. The first objective of this thesis work is then to propose 3D navigation paradigms adapted to the molecular structures of increasing complexity. On the other hand, the interactive context of immersive environments encourages direct interaction with the objects of interest. But the activities of: results collection, simulation and analysis, assume a working environment based on command-line inputs or through specific scripts associated to the tools. Usually, the use of virtual reality is therefore restricted to molecular structures exploration and visualization. The second thesis objective is then to bring all these activities, previously carried out in independent and interactive application contexts, within a homogeneous and unique interactive context. In addition to minimizing the time spent in data management between different work contexts, the aim is also to present, in a joint and simultaneous way, molecular structures and analyses, and allow their manipulation through direct interaction.Our contribution meets these objectives by building on an approach guided by both the content and the task. More precisely, navigation paradigms have been designed taking into account the molecular content, especially geometric properties, and tasks of the expert, to facilitate spatial referencing in molecular complexes and make the exploration of these structures more efficient. In addition, formalizing the nature of molecular data, their analysis and their visual representations, allows to interactively propose analyzes adapted to the nature of the data and create links between the molecular components and associated analyzes. These features go through the construction of a unified and powerful semantic representation making possible the integration of these activities in a unique interactive context.

We present an exploration into realistic locomotion interfaces in video games using spatially convenient input hardware. In particular, we use Nintendo Wii Remotes to create natural mappings between user actions and their representation in a video game. Targeting American Football video games, we used the role of the quarterback as an exemplar since the game player needs to maneuver effectively in a small area, run down the field, and perform evasive gestures such as spinning, jumping, or the "juke". In our study, we developed three locomotion techniques. The first technique used a single Wii Remote, placed anywhere on the user's body, using only the acceleration data. The second technique just used the Wii Remote's infrared sensor and had to be placed on the user's head. The third technique combined a Wii Remote's acceleration and infrared data using a Kalman filter. The Wii Motion Plus was also integrated to add the orientation of the user into the video game. To evaluate the different techniques, we compared them with a cost effective six degree of freedom (6DOF) optical tracker and two Wii Remotes placed on the user's feet. Experiments were performed comparing each to this technique. Finally, a user study was performed to determine if a preference existed among these techniques. The results showed that the second and third technique had the same location accuracy as the cost effective 6DOF tracker, but the first was too inaccurate for video game players. Furthermore, the range of the Wii remote infrared and Motion Plus exceeded the optical tracker of the comparison technique. Finally, the user study showed that video game players preferred the third method over the second, but were split on the use of the Motion Plus when the tasks did not require it.

Navigering i virtuella 3D-miljöer har varit möjligt i många år och sker vanligtvis i samband med till exempel spel och 3D-modellering. Till en persondator används nästan alltid en datormus och ett tangentbord. Musen har visats vara lätthanterlig vid rotering av vyn och kommer inte att vara fokuset i studien. Tangentbordet däremot, som styr vyns position, skulle möjligtvis kunna bytas ut mot något bättre. Vanligtvis används tangenterna W, A, S, och D för förflyttningar, och sedan behövs två tangenter till om det ska vara möjligt att ”flyga” upp och ner. Sex olika tangenter för att styra förflyttningen i tre dimensioner kan vara svårt att lära sig. Trots att vana användare kan hantera det bra, skulle ett mer naturligt sätt att styra på kunna vara enklare för nybörjare, och kanske också för de erfarna. Det begränsade antalet tangenter som används tillåter inte heller finjustering av riktningen. Studien föreslår en alternativ form av 3D-navigering där användaren styr med sitt finger. En Leap Motion kamera ligger på bordet under för att mäta fingrets position, och översätter det till en vektor som kontrollerar vyns hastighet och riktning. Detta är tänkt att vara ett mer naturligt sätt att styra på, då människor har så bra kontroll över sin egen kropp. Utöver det kan även hastigheten justeras genom att dra fingret längre eller kortare sträckor. Vid styrning med tangentbord är justering av hastigheten inte möjligt; undantaget är om användaren kan hålla ner en tangent för att springa, vilket gör att det finns två val av hastigheter. Fingerstyrningen testades och jämfördes direkt mot tangentbordet i ett antal olika experiment. Testerna visar att det går snabbare när tangentbordet används, och ge-nerellt sker färre misstag. När fingerstyrningen används så blir färdsträckorna ofta kortare, speciellt när det krävs mer precision, dock kan detta bero på den lägre has-tigheten som deltagarna hade när de använde fingret. En inmatningsmetod testades bara sju gånger. Under denna period blev fingerstyrningen betydligt snabbare mellan varje försök jämfört med tangentbordet, därför finns anledning att tro att fingerstyr-ningen kommer att förbättras med mer träning. För att få pålitliga resultat skulle en längre studie behöva utföras där deltagarna verkligen hinner lära sig att styra med fingret. Författaren har under utvecklingen av fingerstyrningen blivit snabbare med den än med tangentbordet. Detta är en ytterligare indikation på att det finns potential hos fingerstyrningen som deltagarna aldrig hann uppnå i denna preliminära studie, och att ytterligare experiment krävs.

The purpose of this study was to gain insights of which design guidelines are vital for developing anavigation solution with Augmented Reality from a usability perspective. We wanted to raise awarenessof how the technology could be used to enable both indoor and outdoor navigation.Design Science Research (DSR) was applied as the research methodology, and created prerequisites toidentify problems connected to Luleå University of Technology’s existing map. The study applied the 7steps of the design thinking process, which made it possible to identify the problems. Through surveysand qualitative interviews in combination with observations, it was possible to create an overall picture ofthe problems to understand the users needs. Based on the feedback from the user tests, the prototype couldbe developed and improved.In the second step, 4 themes were set up based on the empirical data collected, which were Navigation,Augmented Reality, General design and Feedback. These themes followed as a common threadthroughout the study and were used to structure the design guidelines and to develop the prototype. Ourstudy shows the importance of throughout the process carry out a consistent design when designing digitalservices. Through this the user can see connections and understand how the application is structured. Toestablish user participation is important at an early stage of the design process, because it contributes togaining insights into deficiency and positive aspects of the prototype, which in turn contributes to meetingthe user needs. One difficulty identified in the empirical data was problems in distinguishing differentfloors in already existing navigation solutions. In order to solve this problem we found it important tocomplement the 2D view in the solution. In the study, this was supplemented by a parallel AR view and atext showing the number of the floor in the 2D view.Keywords: / Denna studie syftade till att skapa förståelse kring vilka designprinciper som utifrån ettanvändbarhetsperspektiv är avgörande när det kommer till utvecklandet av en navigationslösning medAR-funktionalitet. Studien riktade sig mot att öka medvetenhet kring hur tekniken skulle kunna användasför att agera som hjälpmedel i inom-och utomhusnavigation.Med Design Science Research (DSR) som forskningsmetod, skapades förutsättningar för att kunnaidentifiera problematik med Luleå tekniska universitets befintliga karta. I studien tillämpadesdesigntänkande-processens 7 steg, vilket möjliggjorde för att kunna identifiera dessa problem. Genomenkäter och kvalitativa intervjuer i kombination med observationer gick det att skapa en bild överproblematiken kopplat till Luleå tekniska universitets befintliga karta. Utifrån enkäterna skapades även enförståelse över vilka behov som finns utifrån användarnas perspektiv och genom användartester kundeprototypen utvecklas och förbättras.I det andra steget av designtänkande-processen sattes 4 teman upp utifrån studiens empiri, närmarebestämt Navigation, Augmented Reality, Allmän design samt Feedback. Dessa teman följde som en rödtråd genom studien och användes för att strukturera designprinciperna och utforma prototypen. Vår studievisar på vikten av att genomgående genomföra en konsekvent design vid utformandet av digitala tjänster.Genom detta kan användaren se samband och förstå hur applikationen är uppbyggd. Det gick även attkonstatera att användarmedverkan i ett tidigt skede av designprocessen bidrar till att få insikter i bristeroch positiva aspekter av prototypen som i sin tur leder till att kunna uppfylla användarnas behov. Ettproblem som identifierades var svårigheter att urskilja våningsplan i befintliga navigationslösningar. Föratt kunna urskilja våningsplan fann studien det avgörande att hitta ett komplement till 2D-vyn i ennavigationslösning. Inom prototypen kompletterades detta genom en parallell AR-vy samt text i 2D-vyn.

<p>Real-time MRI scanning is used to visualize tissue and organs in motion. The real-time approach requires new interaction techniques to facilitate interaction with the scanning plane. The aim of this thesis is to investigate the use of input with six degrees of freedom – 6DOF. An overview over existing 6DOF input devices is given. Three devices are chosen for implementation and evaluation, Flock of Birds, SpaceBall and SpaceMouse. </p><p>A simulator application is developed to test the different input devices. The simulator purpose is to imitate the real-time scanning situation. To be able to evaluate speed andaccuracy of the different interaction techniques, methods for measuring time and error are developed. A statistical survey is done on two different tasks to gather data of the interaction. The data is analyzed and the result is that the test subjects find the SpaceMouse superior to the other devices thanks to its kinesthetic feed-back properties and ergonomic benefits. However, the statistical data shows that Flock of Birds is the fastest device and no great difference is showed in accuracy between Flock of Birds and SpaceMouse. SpaceBall was the device that generated the least satisfying data.</p>

Real-time MRI scanning is used to visualize tissue and organs in motion. The real-time approach requires new interaction techniques to facilitate interaction with the scanning plane. The aim of this thesis is to investigate the use of input with six degrees of freedom – 6DOF. An overview over existing 6DOF input devices is given. Three devices are chosen for implementation and evaluation, Flock of Birds, SpaceBall and SpaceMouse. A simulator application is developed to test the different input devices. The simulator purpose is to imitate the real-time scanning situation. To be able to evaluate speed andaccuracy of the different interaction techniques, methods for measuring time and error are developed. A statistical survey is done on two different tasks to gather data of the interaction. The data is analyzed and the result is that the test subjects find the SpaceMouse superior to the other devices thanks to its kinesthetic feed-back properties and ergonomic benefits. However, the statistical data shows that Flock of Birds is the fastest device and no great difference is showed in accuracy between Flock of Birds and SpaceMouse. SpaceBall was the device that generated the least satisfying data.

In this thesis I deal with the problem of navigation and autonomous driving using convolutional neural networks. I focus on the main approaches utilizing sensory inputs described in literature and the theory of neural networks, imitation and reinforcement learning. I also discuss the tools and methods applicable to driving systems. I created two deep learning models for autonomous driving in simulated environment. These models use the Dataset Aggregation and Deep Deterministic Policy Gradient algorithms. I tested the created models in the TORCS car racing simulator and compared the result with available sources.