A proposta deste trabalho é contribuir com a construção de uma plataforma de veículo autônomo para viabilizar as pesquisas na área pelo Departamento de Engenharia de Transportes da USP. Até o momento o departamento dispõe de uma plataforma que, a partir de sua posição conhecida, consegue navegar autonomamente até um ponto de destino utilizando apenas uma solução GNSS, no caso, GPS. Para ampliar a mobilidade da plataforma, está sendo sugerida aqui, a adição de sensores inerciais ao veículo, para que ele consiga obter uma solução de posição mesmo em áreas sem cobertura GNSS. Um Sistema de Navegação Inercial não depende de infraestrutura externa, exceto para inicializar suas variáveis, o que neste caso pode ser feito com auxílio de um receptor GPS. Sensores inerciais de alto desempenho são caros, tem alta complexidade mecânica e em geral são de grande porte. A alternativa é o uso de sensores do tipo MEMS que são pequenos, fáceis de serem manipulados e apresentam baixo consumo de energia. A contrapartida é que a solução é mais susceptível a ruído do que seus pares que custam na faixa de centena de milhões de dólares. / The proposal of this paper is to build an autonomous vehicle platform to enable the researches in this area by the Transport Engineering Department of the USP. Until now the Department has a platform that, once its initial position is known, it can navigate autonomously to a destination point using only the GNSS, in this case, GPS. To expand the mobility resources of the platform, it is being suggested here the addition of inertial sensors to the vehicle, enabling it to acquire a position solution even in areas where there is no coverage of the GNSS. An Inertial Navigation System does not depend on an external infra-structure, with the exception on the initial setup, where the GPS can be used to provide this kind of initialization. High performance inertial sensors are expensive, have high mechanical complexity and in general are big. The alternative is the usage of MEMS sensors, which are small, easy to handle and has low power consumption. In the opposite side this solution is more susceptible to noises in comparison to those High performance sensors that cost hundreds of thousands of dollars.

A ground vehicle is a dynamic system containing many non-linear components, ranging from the non-linear engine response to the tyre-road interface. In pursuit of developing driver-assist systems for accident avoidance, as well as fully autonomous vehicles, the application of modern mechatronics systems to vehicles are widely investigated. Extensive work has been done in an attempt to model and control the lateral response of the vehicle system utilising a wide variety of conventional control and intelligent systems theory. The majority of driver models are however intended for low speed applications where the vehicle dynamics are fairly linear. This study proposes the use of adaptive control strategies as robust driver models capable of steering the vehicle without explicit knowledge of vehicle parameters. A Model Predictive Controller (MPC), self-tuning regulator and Linear Quadratic Self-Tuning Regulator (LQSTR) updated through the use of an Auto Regression with eXogenous input (ARX) model that describes the relation between the vehicle steering angle and yaw rate are considered as solutions. The strategies are evaluated by performing a double lane change in simulation using a validated full vehicle model in MSC ADAMS and comparing the maximum stable speed and lateral offset from the required path. It is found that all the adaptive controllers are able to successfully steer the vehicle through the manoeuvre with no prior knowledge of the vehicle parameters. An LQSTR proves to be the best adaptive strategy for driver model applications, delivering a stable response well into the non-linear tyre force regime. This controller is implemented on a fully instrumented Land Rover 110 of the Vehicle Dynamics Group at the University of Pretoria fitted with a semi-active spring-damper suspension that can be switched between two discrete setting representing opposite extremes of the desired response namely: ride mode (soft spring and low damping) and handling mode (stiff spring and high damping). The controller yields a stable response through a severe double lane change (DLC) up to the handling limit of the vehicle, safely completing the DLC at a maximum speed of 90 km/h all suspension configurations. The LQSTR also proves to be robust by following the same path for all suspension configurations through the manoeuvre for vehicle speeds up to 75 km/h. Validation is continued by successfully navigating the Gerotek dynamic handling track, as well as by performing a DLC manoeuvre on an off-road terrain. The study successfully developed and validated a driver model that is robust against variations in vehicle parameters and friction coefficients. / Dissertation (MEng)--University of Pretoria, 2015. / Mechanical and Aeronautical Engineering / Unrestricted

O projeto de mestrado, denominado de forma abreviada como GPUServices, se insere no contexto da pesquisa e do desenvolvimento de métodos de processamento de dados de sensores tridimensionais aplicados a robótica móvel. Tais métodos serão chamados de serviços neste projeto e incluem algoritmos de pré-processamento de nuvens de pontos 3D com segmentação dos dados, a separação e identificação de zonas planares (chão, vias), e detecção de elementos de interesse (bordas, obstáculos). Devido à grande quantidade de dados a serem tratados em um curto espaço de tempo, esses serviços utilizam processamento paralelo por GPU para realizar o processamento parcial ou completo destes dados. A área de aplicação em foco neste projeto visa prover serviços para um sistema ADAS: veículos autônomos e inteligentes, forçando-os a se aproximarem de um sistema de processamento em tempo real devido ao contexto de direção autônoma. Os serviços são divididos em etapas de acordo com a metodologia do projeto, mas sempre buscando a aceleração com o uso de paralelismo inerente: O pré-projeto consiste de organizar um ambiente que seja capaz de coordenar todas as tecnologias utilizadas e que explore o paralelismo; O primeiro serviço tem a responsabilidade de extrair inteligentemente os dados do sensor que foi usado pelo projeto (Sensor laser Velodyne de múltiplos feixes), que se mostra necessário devido à diversos erros de leitura e ao formato de recebimento, fornecendo os dados em uma estrutura matricial; O segundo serviço em cooperação com o anterior corrige a desestabilidade espacial do sensor devido à base de fixação não estar perfeitamente paralela ao chão e devido aos amortecimentos do veículo; O terceiro serviço separa as zonas semânticas do ambiente, como plano do chão, regiões abaixo e acima do chão; O quarto serviço, similar ao anterior, realiza uma pré-segmentação das guias da rua; O quinto serviço realiza uma segmentação de objetos do ambiente, separando-os em blobs; E o sexto serviço utiliza de todos os anteriores para a detecção e segmentação das guias da rua. Os dados recebidos pelo sensor são na forma de uma nuvem de pontos 3D com grande potencial de exploração do paralelismo baseado na localidade das informações. Porém, sua grande dificuldade é a grande taxa de dados recebidos do sensor (em torno de 700.000 pontos/seg.), sendo esta a motivação deste projeto: usar todo o potencial do sensor de forma eficiente ao usar o paralelismo de programação GPU, disponibilizando assim ao usuário serviços de tratamento destes dados. / The master\'s project, abbreviated hence forth as GPUServices, fits in the context of research and development of three-dimensional sensor data processing methods applied to mobile robotics. Such methods will be called services in this project, which include a 3D point cloud preprocessing algorithms with data segmentation, separation and identification of planar areas (ground track), and also detecting elements of interest (borders, barriers). Due to the large amount of data to be processed in a short time, these services should use parallel processing, using the GPU to perform partial or complete processing of these data. The application area in focus in this project aims to provide services for an ADAS system: autonomous and intelligent vehicles, forcing them to get close to a real-time processing system due to the autonomous direction of context.The services are divided into stages according to the project methodology, but always striving for acceleration using inherent parallelism: The pre-project consists of organizing an environment for development that is able to coordinate all used technologies, to exploit parallelism and to be integrated to the system already used by the autonomous car; The first service has a responsibility to intelligently extract sensor data that will be used by the project (Laser sensor Velodyne multi-beam), it appears necessary because of the many reading errors and the receiving data format, hence providing data in a matrix structure; The second service, in cooperation with the above, corrects the spatial destabilization due to the sensor fixing base not perfectly parallel to the ground and due to the damping of the vehicle; The third service separates the environment into semantics areas such as ground plane and regions below and above the ground; The fourth service, similar to the above, performs a pre-segmentation of street cruds; The fifth service performs an environmental objects segmentation, separating them into blobs; The sixth service uses all prior to detection and segmentation of street guides.The received sensor data is structured in the form of a cloud of points. They allow processing with great potential for exploitation of parallelism based on the location of the information. However, its major difficulty is the high rate of data received from the sensor (around 700,000 points/sec), and this gives the motivation of this project: to use the full potential of sensor to efficiently use the parallelism of GPU programming, therefore providing data processing services to the user, providing services that helps and make the implementation of ADAS systems easier and/or faster.

Tato práce se zabývá simulačními prostředky pro vývoj algoritmů pro řízení autonomních automobilů. V zásadě lze rozdělit na dvě části, na rešeršní, teoretickou, a praktickou, vývojovou. V té prvně zmíněné je uveden přehled dostupných nástrojů pro simulaci autonomních vozidel, jedná se jak o nástroje open-sourcové tak placené. Dále se v teoretické části popisuje princip a nástroje, resp. enginy pro řešení dynamických rovnic na počítači. Důraz je kladen na fyzikální engine Box2D který je dle zadání této práce využit ve druhé části teze pro vývoj vlastního prostředí simulujícího autonomní automobil.

abstract: As robots become mechanically more capable, they are going to be more and more integrated into our daily lives. Over time, human’s expectation of what the robot capabilities are is getting higher. Therefore, it can be conjectured that often robots will not act as human commanders intended them to do. That is, the users of the robots may have a different point of view from the one the robots do. The first part of this dissertation covers methods that resolve some instances of this mismatch when the mission requirements are expressed in Linear Temporal Logic (LTL) for handling coverage, sequencing, conditions and avoidance. That is, the following general questions are addressed: * What cause of the given mission is unrealizable? * Is there any other feasible mission that is close to the given one? In order to answer these questions, the LTL Revision Problem is applied and it is formulated as a graph search problem. It is shown that in general the problem is NP-Complete. Hence, it is proved that the heuristic algorihtm has 2-approximation bound in some cases. This problem, then, is extended to two different versions: one is for the weighted transition system and another is for the specification under quantitative preference. Next, a follow up question is addressed: * How can an LTL specified mission be scaled up to multiple robots operating in confined environments? The Cooperative Multi-agent Planning Problem is addressed by borrowing a technique from cooperative pathfinding problems in discrete grid environments. Since centralized planning for multi-robot systems is computationally challenging and easily results in state space explosion, a distributed planning approach is provided through agent coupling and de-coupling. In addition, in order to make such robot missions work in the real world, robots should take actions in the continuous physical world. Hence, in the second part of this thesis, the resulting motion planning problems is addressed for non-holonomic robots. That is, it is devoted to autonomous vehicles’ motion planning in challenging environments such as rural, semi-structured roads. This planning problem is solved with an on-the-fly hierarchical approach, using a pre-computed lattice planner. It is also proved that the proposed algorithm guarantees resolution-completeness in such demanding environments. Finally, possible extensions are discussed. / Dissertation/Thesis / Doctoral Dissertation Computer Science 2019

This thesis is carried out in the context of Advanced Driver Assistance Systems, and especially autonomous vehicles. Its aim is to propose a method to enhance localization of vehicles on roads. I suggests using a camera to detect lane markings, and to match these to a map to extract the corrected position of the vehicle.The thesis is divided in three parts dealing with: the map, the line detector and the evaluation. The map is based on the OpenStreetMap data. The line detector is a based on ridge detection. The results are compared with an Iterative Closest Point algorithm. It also focuses on implementing the components under a real-time constraint. Technologies such as ROS, for synchronization of the data, and CUDA, for parallelization, are used.

The aim of this report is to provide an analysis of potential benefits of a fleet of Shared Autonomous Vehicles (SAV) providing a taxi service to replace private car commuter trips in a metropolitan area. We develop a framework for dynamic allocation of SAVs to passenger trips, empty-vehicle routing and multi-criteria evaluation with regard to passenger waiting time, trip time and fleet size. Using a representation of current private trip demand for the Stockholm metropolitan area and a detailed road network representation, different scenarios (varying levels of accepted passenger waiting time at origin and accepted increase in travel time) are compared with respect to passenger travel time, number of vehicles needed and vehicle mileage. In a second step the environmental impacts of the different scenarios are assessed and compared using a life cycle approach. The assessment includes both a fleet consisting of currently in use gasoline and diesel cars as well as electrical cars. The results show that an SAV-based personal transport system has the potential to provide an on-demand door-to-door transport with a high level of service, using less than 10 % of today's private cars and parking places. In order to provide an environmental benefit and lower congestion an SAV-based personal transport system requires users to accept ride-sharing, allowing a maximum 30% increase of their travel time (15% on average) and a start time window of 10 minutes. In a scenario where users are not inclined to accept any reduced level of service, i.e. no ride-sharing and no delay, empty vehicle drive of an SAV-based personal transport system will lead to increased road traffic increasing environmental impacts and congestion. Envisioning a future using electrical cars a SAV-based system and electrical vehicle technology seems to be a “perfect” match that could definitely contribute to a sustainable transport system in Stockholm.

Autonomous vehicle technology has seen rapid development thanks to advances in artificial intelligence. Among the various sectors, agriculture is one sector that is testing the potential of autonomous vehicle robots to meet the growing demands of society. Cutting or "Mowing" grass is one potential application that can be automated with AI-driven vehicles on large farms to increase efficiency. However, the increasing reliance on artificial intelligence models for decision-making, such as for navigation, raises the question of how safe these models are and how we can assess the safety of such algorithms. As the safety of AI is still an open challenge, very little research has addressed this problem, and even less in the field of agriculture. The aim of this work is to develop a framework for evaluating the safety of AI perception models used in autonomous vehicle robots in agriculture. The proposed methodology evaluates safety in three main stages: sub-system level, system-level, and post-deployment, along with a preliminary stage for defining boundaries. The feasibility of the framework was also tested on an AI perception system present in a prototype autonomous mowing vehicle to identify areas of safety concern. / Autonom fordonsteknologi har haft en snabb utveckling tack vare framstegen inom konstgjord intelligens. Bland de olika sektorerna är jordbruk en sektor som testar potentialen för autonoma fordonsrobotar för att möta samhällets växande krav. Skär eller "Mowing" gräs är en potentiell applikation som kan automatiseras med AI-drivna fordon på stora gårdar för att öka effektiviteten. Det ökande beroendet av modeller för konstgjord intelligens för beslutsfattande, till exempel för navigering, väcker emellertid frågan om hur säkra dessa modeller är och hur vi kan bedöma säkerheten för sådana algoritmer. Eftersom AI: s säkerhet fortfarande är en öppen utmaning har mycket lite forskning tagit upp detta problem och ännu mindre inom jordbruksområdet. Syftet med detta arbete är att utveckla en ram för utvärdering av säkerheten för AI-uppfattningsmodeller som används i autonoma fordonsrobotar i jordbruket. Den föreslagna metodologin utvärderar säkerheten i tre huvudsteg: delsystemnivå, systemnivå och efterutplacering, tillsammans med ett preliminärt steg för att definiera gränser. Ramens genomförbarhet testades också på ett AI-uppfattningssystem som finns i ett prototyp autonomt klippfordon för att identifiera områden med säkerhetsproblem.

<p>Unsafe lane changes have been identified as a common factor in motor vehicle accidents. It would be helpful, particularly for automated vehicles, to know if there are behaviors of vehicles, beyond a directional signal, or characteristics of the traffic environment that correlated with a higher probability of an unsafe lane change (lane changes without a directional signal). This work investigates what the observable cues are that drivers use to determine the relative safety when overtaking front vehicles, and if drivers make more lane changes under certain conditions on highways. This study utilizes interviews, surveys, 3D animation software, and highway driving public footage for data collection and experiments. It is found that a side-to-side motion of the front vehicle or a factor that might trigger a side-to-side motion of the front vehicle in the environment is the key marker that indicates a possible unsafe lane change, and it is also found that traffic speed, time of day, traffic flow, and a combination of traffic density & number of lanes & vehicle count all have effects on drive’s decision on making lane changes on different levels.</p>

Object detection exists in many countries around the world after a recent growing interest for autonomous vehicles in the last decade. This paper focuses on a vision-based approach focusing on vehicles and pedestrians detection in real-time as a perception for autonomous vehicles, using a convolutional neural network for object detection. A developed YOLOv3-tiny model is trained with the INRIA dataset to detect vehicles and pedestrians, and the model also measures the distance to the detected objects. The machine learning process is leveraged to describe each step of the training process, it also combats overfitting and increases the speed and accuracy. The authors were able to increase the mean average precision; a way to measure accuracy for object detectors; 31.3\% to 62.14\% based on the result of the training that was done. Whilst maintaining a speed of 18 frames per second.