Indiana University-Purdue University Indianapolis (IUPUI) / Advancement in modern technology has brought with it an advent of increased interest in self-driving. The rapid growth in interest has caused a surge in the development of autonomous vehicles which in turn brought with itself a few challenges. To overcome these new challenges, automotive companies are forced to invest heavily in the research and development of autonomous vehicles. To overcome this challenge, simulations are a great tool in any arsenal that’s inclined towards making progress towards a self-driving autonomous future. There is a massive growth in the amount of computing power in today’s world and with the help of the same computing power, simulations will help test and simulate scenarios to have real time results. However, the challenge does not end here, there is a much bigger hurdle caused by the growing complexities of modelling a complete simulation environment. This thesis focuses on providing a solution for modelling a RADAR sensor for a simulation environment. This research presents a RADAR modeling technique suitable for autonomous vehicle simulation environment using open-source utilities. This study proposes to customize an onboard LiDAR model to the specification of a desired RADAR field of view, resolution, and range and then utilizes a density-based clustering algorithm to generate the RADAR output on an open-source graphical engine such as Unreal Engine (UE). High fidelity RADAR models have recently been developed for proprietary simulation platforms such as MATLAB under its automated driving toolbox. However, open-source RADAR models for open-source simulation platform such as UE are not available. This research focuses on developing a RADAR model on UE using blueprint visual scripting for off-road vehicles. The model discussed in the thesis uses 3D pointcloud data generated from the simulation environment and then clipping the data according to the FOV of the RADAR specification, it clusters the points generated from an object using DBSCAN. The model gives the distance and azimuth to the object from the RADAR sensor in 2D. This model offers the developers a base to build upon and help them develop and test autonomous control algorithms requiring RADAR sensor data. Preliminary simulation results show promise for the proposed RADAR model.

<div>We describe a new task called The Amazing Race: Robot Edition. In this task, the robot is placed in a real, unknown environment, without a map, and asked to find a designated location. It will need to explore its surroundings, find and approach people, engage them in a dialogue to obtain directions to the goal, and follow those directions to the hallway with the goal. We describe and implement a variety of robotic behaviors that performs each of these functions. We test these in the real world in test environments that were distinct from the training environments where we developed our methods and trained our models. Additionally, these test environments were completely unmodified and reflect the state of the real world.</div><div>First, we describe how our robotic system solves this problem where the environment is constrained to a single floor or a single building. We demonstrate that we are able to find a goal location in never-before-seen environments. Next, we describe a machine-learned approach to the dialogue and components of our system to make it more robust to the diversity and noisiness of navigational instructions someone may provide.</div>

Robust and timely perception of the environment is an essential requirement of all autonomous and semi-autonomous systems. This necessity has been the main factor behind the rapid growth and adoption of LiDAR sensors within the ADAS sensor suite. In this thesis, we develop a fast and accurate 3D object detector that converts raw point clouds collected by LiDARs into sparse occupancy cuboids to detect cars and other road users using deep convolutional neural networks. The proposed pipeline reduces the runtime of PointPillars by 43% and performs on par with other state-of-the-art models. We do not gain improvements in speed by compromising the network's complexity and learning capacity but rather through the use of an efficient input encoding procedure. In addition to rigorous profiling on three different platforms, we conduct a comprehensive error analysis and recognize principal sources of error among the predicted attributes. Even though point clouds adequately capture the 3D structure of the physical world, they lack the rich texture information present in color images. In light of this, we explore the possibility of fusing the two modalities with the intent of improving detection accuracy. We present a late fusion strategy that merges the classification head of our LiDAR-based object detector with semantic segmentation maps inferred from images. Extensive experiments on the KITTI 3D object detection benchmark demonstrate the validity of the proposed fusion scheme.

Neuromorphic computing has many opportunities in future autonomous systems, especially those that will operate at the edge. However, there are relatively few demonstrations of neuromorphic implementations on real-world applications, partly because of the lack of availability of neuromorphic hardware and software, but also because of the lack of availability of an accessible demonstration platform. In this work, we propose utilizing the F1Tenth platform as an evaluation task for neuromorphic computing. F1Tenth is a competition wherein one tenth scale cars compete in an autonomous racing task; there are significant open source resources in both software and hardware for realizing this task. We present a workflow with neuromorphic hardware, software, and training that can be used to develop a spiking neural network for neuromorphic hardware deployment to perform autonomous racing. We present initial results on utilizing this approach for this small-scale, real-world autonomous vehicle task.

abstract: With the growth of autonomous vehicles’ prevalence, it is important to understand the relationship between autonomous vehicles and the other drivers around them. More specifically, how does one’s knowledge about autonomous vehicles (AV) affect positive and negative affect towards driving in their presence? Furthermore, how does trust of autonomous vehicles correlate with those emotions? These questions were addressed by conducting a survey to measure participant’s positive affect, negative affect, and trust when driving in the presence of autonomous vehicles. Participants’ were issued a pretest measuring existing knowledge of autonomous vehicles, followed by measures of affect and trust. After completing this pre-test portion of the study, participants were given information about how autonomous vehicles work, and were then presented with a posttest identical to the pretest. The educational intervention had no effect on positive or negative affect, though there was a positive relationship between positive affect and trust and a negative relationship between negative affect and trust. These findings will be used to inform future research endeavors researching trust and autonomous vehicles using a test bed developed at Arizona State University. This test bed allows for researchers to examine the behavior of multiple participants at the same time and include autonomous vehicles in studies. / Dissertation/Thesis / Masters Thesis Human Systems Engineering 2019

No description available.

A lot of features that need to be scanned underground should not be disturbed, from waterlines to unmarked graves. A non-invasive way of probing underground is Ground-Penetrating Radar (GPR). GPR finds differences in materials with radar waves. However, GPR is human-operated and its position is generally determined with a GPS. In some cases, the presence of a human operator can be dangerous, and in other cases, the GPS is not reliable (i.e. mines, glaciers). Therefore there are situations where an autonomous and non-GPS-reliant solution is preferable. The current state of the autonomous GPR system targeted in this work has a non-intuitive GUI that requires an experienced hand to operate. I present an updated hardware and software platform with an intuitive GUI. This updated autonomous system continuously builds a map of its surroundings with Simultaneous Mapping And Localization (SLAM). SLAM localizes itself within the map through sensor-fused position estimates. After the survey is completed the positions are saved and integrated with radar data to be visualized. Robot Operating System 2 (ROS2) is the software I used that allows communications between hardware components, software systems, and the GUI. The new hardware package uses only one source of power and is built using quick connectors that allow for quick removal from the GPR platform. This system allows for intuitive autonomous survey planning and execution in any field paired with a simple way of visualizing data.

No description available.

Self-driving is a technology that has been envisioned in science fiction movies or in speculative design for quite some time. However, it is one of the few future technologies that is relatively easy to imagine, but very difficult to implement it into reality due to complexity coming from variability in AI. This discrepancy between reality and imagination is what makes achieving trust in self-driving vehicles more challenging, especially regarding the fact that driving is regarded as a daily task for some people. Keeping into consideration how most of the other projects done to enhance trust in automation deals with full automation, this thesis focuses how trust can be defined in semi autonomous vehicles. This middle ground setting with humans and AI systems working together needs more factors to be considered to make it autonomous, at the same time requiring a higher level of trust from drivers. An additional layer of a takeover situation from driver to AI and vice versa in a semi-autonomous setting would require more level of trust than a full self-driving vehicle where drivers do not have to control anything.Volvo Cars, an automobile manufacturer brand that has its strong focus on safety, was collaborated with in this project to support developing a notion of trust in autonomous systems. The purpose of this collaboration with Volvo Cars was to receive support in any expert knowledge in the mobility field and to create a project that is relevant to the current development state and future vision of autonomous vehicles. In order to provide an environment where drivers can calibrate trust inside vehicles, FiDO, a tangible driving assistant for building trust, was designed through a participatory design process. FiDO provides an environment for setting mutual expectation between driver and vehicle through communicating vehicle’s status and driver’s feedback with poetic visuals. FiDO learns from driver’s behaviors and their direct feedback, which provides personalized content and autonomous driving as an outcome of learning. FiDO’s usage can be adjusted based on driver’s trust level and characteristics of the service of where automation technology is used.This thesis does not cover the entire notion of trust in automation, but focuses particularly on building trust from a driver’s point of view. With including users throughout the process, this is a proof of concept how automation technology and notion of trust can be built with driver’s participation. Although detailed technological feasibility of including both humans and AI in one place to build an autonomous system were not considered into practical levels, this thesis emphasizes how we can also establish trust voluntarily from a user’s point of view.

The feasibility of neural networks to control dynamic systems is examined. Control of a one-dimensional problem is initially investigated to develop an understanding of the structure and simulation of the neural networks. A nondimensional problem is also explored to apply a single neural network design to controlling a class of systems with a wide variety of modeling parameters. Finally, these techniques are applied to control a space vehicle to transfer, intercept, and rendezvous with another orbiting vehicle using the Clohessy-Wiltshire equations of relative motion in two dimensions. A combination of open-loop and closed-loop neural network controllers is shown to work effectively for this problem. Noise is added to the neural network inputs to demonstrate the robustness of these networks. / Ph. D.