Autonomous vehicles perform semantic segmentation to orient themselves, but training neural networks for semantic segmentation requires large amounts of labeled data. A hand-labeled real-life dataset requires considerable effort to create, so we instead turn to virtual simulators where the segmented labels are known to generate large datasets virtually for free. This work investigates how effective synthetic datasets are in driving scenarios by collecting a dataset from a simulator and testing it against a real-life hand-labeled dataset. We show that we can get a model up and running faster by mixing synthetic and real-life data than traditional dataset collection methods and achieve close to baseline performance.

abstract: Autonomous Vehicles (AVs), or self-driving cars, are poised to have an enormous impact on the automotive industry and road transportation. While advances have been made towards the development of safe, competent autonomous vehicles, there has been inadequate attention to the control of autonomous vehicles in unanticipated situations, such as imminent crashes. Even if autonomous vehicles follow all safety measures, accidents are inevitable, and humans must trust autonomous vehicles to respond appropriately in such scenarios. It is not plausible to program autonomous vehicles with a set of rules to tackle every possible crash scenario. Instead, a possible approach is to align their decision-making capabilities with the moral priorities, values, and social motivations of trustworthy human drivers.Toward this end, this thesis contributes a simulation framework for collecting, analyzing, and replicating human driving behaviors in a variety of scenarios, including imminent crashes. Four driving scenarios in an urban traffic environment were designed in the CARLA driving simulator platform, in which simulated cars can either drive autonomously or be driven by a user via a steering wheel and pedals. These included three unavoidable crash scenarios, representing classic trolley-problem ethical dilemmas, and a scenario in which a car must be driven through a school zone, in order to examine driver prioritization of reaching a destination versus ensuring safety. Sample human driving data in CARLA was logged from the simulated car’s sensors, including the LiDAR, IMU and camera. In order to reproduce human driving behaviors in a simulated vehicle, it is necessary for the AV to be able to identify objects in the environment and evaluate the volume of their bounding boxes for prediction and planning. An object detection method was used that processes LiDAR point cloud data using the PointNet neural network architecture, analyzes RGB images via transfer learning using the Xception convolutional neural network architecture, and fuses the outputs of these two networks. This method was trained and tested on both the KITTI Vision Benchmark Suite dataset and a virtual dataset exclusively generated from CARLA. When applied to the KITTI dataset, the object detection method achieved an average classification accuracy of 96.72% and an average Intersection over Union (IoU) of 0.72, where the IoU metric compares predicted bounding boxes to those used for training. / Dissertation/Thesis / Masters Thesis Mechanical Engineering 2020

No description available.

Navigating through road work zones remains a challenge for the development of autonomous driving technology. While HD maps are essential for accurate local- ization and navigation in autonomous vehicles, they face issues when encountering dynamic and constantly changing situations on the road such as road construction sites. As a result, autonomous vehicles need to rely on their onboard sensor data for safe navigation through construction zones. This thesis focuses on low-level fusion-based methods, using both point cloud and image data for the detection of road construction sites. The primary objective is to identify temporary traffic control devices like delineator posts, safety barriers, and traffic cones which play a role in ensuring road safety while maintaining smooth traffic flow throughout construction areas. To achieve this, the CARLA simulator is used to generate an autonomous driving dataset that represents various road construction sites that are frequently observed in German regions. This dataset forms the basis for evaluating four state-of-the- art low-level LiDAR-camera fusion-based methods. By establishing a benchmark, this thesis presents a proof of concept for successful road work zone detection. The results demonstrate the effectiveness of low-level fusion-based methods in identifying road construction sites and open the door for further developments, emphasizing its potential impact on advancing autonomous driving technology within work zones.

Autonomous navigation is a critical component in the development of self-driving vehicles. This thesis explores the application of deep reinforcement learning (DRL) for autonomous navigation within the CARLA simulator, an open-source simulation plat form designed for autonomous driving research. The work focuses on training agents to make optimal driving decisions in dynamic urban environments without human inter vention. Deep learning models were combined with reinforcement learning techniques so the vehicle could perceive its surroundings, predict outcomes, and take appropriate actions to navigate safely. The study evaluates the performance of a state-of-the-art DRL algorithm, Proxi mal policy optimization (PPO), while actively addressing and overcoming challenges like sparse rewards, training stability, and generalization to unseen scenarios. A cus tom reward function was crafted to prioritize collision avoidance, lane-keeping, smooth acceleration, and steering, ensuring the agent adheres to realistic driving behavior. Experimental results demonstrated that the DRL-based agent achieved promising per formance in various simulated driving tasks, including maintaining speed, following traffic signals, lane-following, and intersection handling. Furthermore, the agent ex hibited commendable performance in novel environments, highlighting its capacity to generalize and adapt efficiently. This thesis contributes to the understanding of integrating DRL for autonomous navigation in simulation-based environments and highlights the CARLA simulator’s role as a robust testing ground. The findings lay the groundwork for further ad vancements in sim-to-real transfer and scalable training methods for autonomous vehicles.:1 Introduction 1.1 Problem Statement 5 1.2 Thesis Structure 6 2 Background 2.1 Machine Learning 7 2.2 Deep Learning 8 2.2.1 Feed-Forward Network 9 2.3 Reinforcement Learning 10 2.3.1 Markov Decision Process 10 2.3.2 Bellman Equation 11 2.3.3 Reward Function 13 2.3.4 Action Spaces 13 2.4 Deep Reinforcement Learning 15 2.4.1 Policy-Based Approaches 15 2.4.2 Proximal Policy Optimization 17 3 Experiment Setup 3.1 CARLA Simulator 21 3.1.1 Vehicle Control 22 3.1.2 Maps 23 3.1.3 Waypoints and Routes 24 3.2 Environment Setup 25 3.3 Deep Reinforcement Learning Setup 26 3.3.1 State Space 26 3.3.2 Action Space 29 3.3.3 Reward Function 30 3.4 Network Architecture 33 3.5 Model Training 34 4 Evaluation 4.1 Evaluating Agent on New Maps 38 5 Conclusion