Search Results

This paper focuses on finding and analyzing the relevant parameters affecting traffic flow when autonomous vehicles are introduced for ride hailing applications and autonomous shuttles are introduced for circulator applications in geo-fenced urban areas. For this purpose, different scenarios have been created in traffic simulation software that model the different levels of autonomy, traffic density, routes, and other traffic elements. Similarly, software that specializes in vehicle dynamics, physical limitations, and vehicle control has been used to closely simulate realistic autonomous vehicle behavior under such scenarios. Different simulation tools for realistic autonomous vehicle simulation and traffic simulation have been merged together in this paper, creating a realistic simulator with Hardware-in-the-Loop (HiL), Traffic-in-the-Loop (TiL), and Software in-the-Loop (SiL) simulation capabilities.

Autonomous Vehicles (AVs) operate based on image information and 3D maps generated by sensors like cameras, LIDARs and RADARs. This information is processed by the on-board processing units to provide the right actuation signals to drive the vehicle. For safe operation, these sensors should provide continuous high quality data to the processing units without interruption in all driving conditions like dust, rain, snow and any other adverse driving conditions. Any contamination on the sensor surface/lens due to rain droplets, snow, and other debris would result in adverse impact to the quality of data provided for sensor fusion and this could result in error states for autonomous driving. In particular, snow is a common contamination condition during driving that might block a sensor surface or camera lens. Predicting and preventing snow accumulation over the sensor surface of an AV is important to overcome this challenge.

The transportation sector is facing three revolutions: shared mobility, electrification, and autonomous driving. To inform decision making and guide smart transportation system development at the city-level, it is critical to model and evaluate how travelers will behave in these systems. Two key components in such models are (1) individual travel demands with high spatial and temporal resolutions, and (2) travelers’ sociodemographic information and trip purposes. These components impact one’s acceptance of autonomous vehicles, adoption of electric vehicles, and participation in shared mobility. Existing methods of travel demand generation either lack travelers’ demographics and trip purposes, or only generate trips at a zonal level. Higher resolution demand and sociodemographic data can enable analysis of trips’ shareability for car sharing and ride pooling and evaluation of electric vehicles’ charging needs.

With the rapid development of artificial intelligence, the autonomous vehicles (AV) have attracted considerable attention in the automotive industry. However, different factors negatively impact the adoption of the AVs, delaying their successful commercialization. Accretion of atmospheric icing, especially wet snow, on AV sensors causes blockage on their lenses, making them prone to lose their sight, in turn, increasing potential chances of accidents. In this study, two different designs are proposed in order to prevent snow accretion on the lenses of AVs via air flow across the lens surface. In both designs, lenses made of plain glass and superhydrophobic coated glass surfaces are tested. While some researchers have shown promise of water repellency on superhydrophobic surfaces, more snow accretion is observed on the superhydrophobic surfaces, when compared to the plain glass lenses.

Successful demonstrations of fully autonomous vehicle operation in controlled situations are leading to increased research investment and activity. This has already resulted in significant advancements in the underlying technologies necessary to make it a practical reality someday. Not only are these idealized events sparking imaginations with the potential benefits for safety, convenience, fuel economy and emissions, they also embolden some to make somewhat surprising and sometimes astonishing projections for their appearance on public roads in the near future. Are we now ready for a giant leap forward to the self-driving car with all its complexity and inter-dependencies? Humans will need to grow with and adapt to the technological advancements of the machine and we'll deeply challenge our social and political paradigms before we're done. Even if we as engineers are ready, is the driving public ready?

In this work, we outline a process for traffic light detection in the context of autonomous vehicles and driver assistance technology features. For our approach, we leverage the automatic annotations from virtually generated data of road scenes. Using the automatically generated bounding boxes around the illuminated traffic lights themselves, we trained an 8-layer deep neural network, without pre-training, for classification of traffic light signals (green, amber, red). After training on virtual data, we tested the network on real world data collected from a forward facing camera on a vehicle. Our new region proposal technique uses color space conversion and contour extraction to identify candidate regions to feed to the deep neural network classifier. Depending on time of day, we convert our RGB images in order to more accurately extract the appropriate regions of interest and filter them based on color, shape and size. These candidate regions are fed to a deep neural network.

While operating a vehicle in either autonomous or occupant piloted mode, an array of sensors can be used to guide the vehicle including stereo cameras. The state-of-the-art distance map estimation algorithms, e.g. stereo matching, usually detect corresponding features in stereo images, and estimate disparities to compute the distance map in a scene. However, depending on the image size, content and quality, the feature extraction process can become inaccurate, unstable and slow. In contrast, we employ deep convolutional neural networks, and propose two architectures to estimate distance maps from stereo images. The first architecture is a simple and generic network that identifies which features to extract, and how to combine them in a multi-resolution framework. The second architecture is a more specialized one that extracts local similarity information from two images, which are used for stereo feature matching, and fuses them at multiple resolutions to generate the distance map.

Software for autonomous vehicles is highly complex and requires vast amount of vehicle testing to achieve a certain level of confidence in safety, quality and reliability. According to the RAND Corporation, a 100 vehicle fleet running 24 hours a day 365 days a year at a speed of 40 km/hr, would require 17 billion driven kilometers of testing and take 518 years to fully validate the software with 95% confidence such that its failure rate would be 20% better than the current human driver fatality rate [1]. In order to reduce cost and time to accelerate autonomous software development, Hardware-in-the-Loop (HIL) simulation is used to supplement vehicle testing. For autonomous vehicles, path following controls are an integral part for achieving lateral control. Combining the aforementioned concepts, this paper focuses on a real-time implementation of a path-following lateral controller, developed by Freund and Mayr [2].

Nowadays, the automotive industry is experiencing the advent of unprecedented applications with connected devices, such as identifying safe users for insurance companies or assessing vehicle health. To enable such applications, driving behavior data are collected from vehicles and provided to third parties (e.g., insurance firms, car sharing businesses, healthcare providers). In the new wave of IoT (Internet of Things), driving statistics and users’ data generated from wearable devices can be exploited to better assess driving behaviors and construct driver models. We propose a framework for securely collecting data from multiple sources (e.g., vehicles and brought-in devices) and integrating them in the cloud to enable next-generation services with guaranteed user privacy protection.

This paper proposes a low-cost but indirect method for occupancy detection and occupant counting purpose in current and future automotive systems. It can serve as either a way to determine the number of occupants riding inside a car or a way to complement the other devices in determining the occupancy. The proposed method is useful for various mobility applications including car rental, fleet management, taxi, car sharing, occupancy in autonomous vehicles, etc. It utilizes existing on-board motion sensor measurements, such as those used in the vehicle stability control function, together with door open and closed status. The vehicle’s motion signature in response to an occupant’s boarding and alighting is first extracted from the motion sensors that measure the responses of the vehicle body. Then the weights of the occupants are estimated by fitting the vehicle responses with a transient vehicle dynamics model.

This paper presents two brake control functions which are initiated when there is an impact force applied to a host vehicle. The impact force is generated due to the host vehicle being collided with or by another vehicle or object. The first function - called the post-impact braking assist - initiates emergency brake assistance if the driver is braking during or right after the collision. The second function - called the post-impact braking - initiates autonomous braking up to the level of the anti-lock-brake system if the driver is not braking during or right after the collision. Both functions intend to enhance the current driver assistance features such as emergency brake assistance, electronic stability control, anti-brake-lock system, collision mitigation system, etc.

Performance testing and evaluation always plays an important role in the developmental process of a vehicle, which also applies to autonomous vehicles. The complex nature of an autonomous vehicle from architecture to functionality demands even more quality-and-quantity controlled testing and evaluation than ever before. Most of the existing testing methodologies are task-or-scenario based and can only support single or partial functional testing. These approaches may be helpful at the initial stage of autonomous vehicle development. However, as the integrated autonomous system gets mature, these approaches fall short of supporting comprehensive performance evaluation. This paper proposes a novel hierarchical and systematic testing and evaluation approach to bridge the above-mentioned gap.

This paper summarizes the development of a wireless message from infrastructure-to-vehicle (I2V) for safety applications based on Dedicated Short-Range Communications (DSRC) under a cooperative agreement between the Crash Avoidance Metrics Partners LLC (CAMP) and the Federal Highway Administration (FHWA). During the development of the Curve Speed Warning (CSW) and Reduced Speed Zone Warning with Lane Closure (RSZW/LC) safety applications [1], the Basic Information Message (BIM) was developed to wirelessly transmit infrastructure-centric information. The Traveler Information Message (TIM) structure, as described in the SAE J2735, provides a mechanism for the infrastructure to issue and display in-vehicle signage of various types of advisory and road sign information. This approach, though effective in communicating traffic advisories, is limited by the type of information that can be broadcast from infrastructures.

The automotive industry is facing many significant challenges that go far beyond the design and manufacturing of automobile products. Connected, autonomous and electric vehicles, smart cities, urbanization and the car sharing economy all present challenges in a fast-changing environment which the automotive industry must adapt to. Cars no longer are just standalone systems, but have become constituent systems (CS) in larger System of Systems (SoS) context. This is reflected in the emergence of several acronyms such as vehicle-to-everything (V2X), vehicle-to-vehicle (V2V), vehicle-to-infrastructure (V2I), and vehicle-to-grid (V2G) expressions. System of Systems are defined systems of interest whose elements (constituent systems) are managerially and operationally independent systems. This interoperating and/or integrated collection of constituent systems usually produce results unachievable by the individual systems alone, for example the use of car batteries as virtual power plants.

This paper takes a look at one of the problems associated with the concept of autonomous control of vehicles in the current traffic environment, namely the changing of lanes. Given the increase in traffic density on highways and interstate roads over the past few decades, safe navigation of individual vehicles has required increased driver attention and diligence to an increased number of visual information cues. The concept of autonomous vehicles operating without driver intervention in the present traffic system appears daunting. One aspect of traffic maneuvering involves changing lanes to a position between two other vehicles. Although this aspect appears straightforward, it is the lack of accurate knowledge of other vehicle maneuvering which makes the task difficult. Using Model Predictive Control (MPC) techniques, the task is addressed in an optimization problem framework.

Traditionally, the controls system in production vehicles with automatic transmission interprets the driver’s accelerator pedal position as a demand for transmission input torque. However, with the advent of electrified vehicles, where actuators are located at different positions in the drivetrain, and of autonomous vehicles, which are self-driving, it is more convenient to interpret the demand (either human or virtual) in vehicle acceleration or wheel torque domain. To this end, a Wheel Torque-based longitudinal Control (WTC) framework was developed, wherein demands can be converted accurately between the vehicle acceleration or wheel torque domain and the transmission assembly input torque domain.

With the proliferation of ADAS and autonomous systems, the quality and quantity of the data to be used by vehicles has become crucial. In-vehicle sensors are evolving, but their usability is limited to their field of view and detection distance. V2X communication systems solve these issues by creating a cooperative perception domain amongst road users and the infrastructure by communicating accurate, real-time information. In this paper, we propose a novel Consolidated Object Data Service (CODS) for multi-Radio Access Technology (RAT) V2X communication. This service collects information using BSM packets from the vehicular network and perception information from infrastructure-based sensors. The service then fuses the collected data, offering the communication participants with a consolidated, deduplicated, and accurate object database. Since fusing the objects is resource intensive, this service can save in-vehicle computation costs.

To enable smooth and low-risk autonomous driving in the presence of other road users, such as cyclists and pedestrians, appropriate predictive safe speed control strategies relying on accurate and robust prediction models should be employed. However, difficulties related to driving scene understanding and a wide variety of features influencing decisions of other road users significantly complexifies prediction tasks and related controls. This paper proposes a hierarchical neural network (NN)-based prediction model of pedestrian crossing behavior, which is aimed to be applied within an autonomous vehicle (AV) safe speed control strategy. Additionally, different single-level prediction models are presented and analyzed as well, to serve as baseline approaches.

This paper describes a multi-sensor data set, suitable for testing algorithms to detect and track pedestrians and cyclists, with an autonomous vehicle’s sensor suite. The data set can be used to evaluate the benefit of fused sensing algorithms, and provides ground truth trajectories of pedestrians, cyclists, and other vehicles for objective evaluation of track accuracy. One of the principal bottlenecks for sensing and perception algorithm development is the ability to evaluate tracking algorithms against ground truth data. By ground truth we mean independent knowledge of the position, size, speed, heading, and class of objects of interest in complex operational environments. Our goal was to execute a data collection campaign at an urban test track in which trajectories of moving objects of interest are measured with auxiliary instrumentation, in conjunction with several autonomous vehicles (AV) with a full sensor suite of radar, lidar, and cameras.

This manuscript compares window-based data imputation approaches for data coming from connected vehicles during actual driving scenarios and obtained using on-board data acquisition devices. Three distinct window-based approaches were used for cleansing and imputing the missing values in different CAN-bus (Controller Area Network) signals. Lengths of windows used for data imputation for the three approaches were: 1) entire time-course for each vehicle ID, 2) day, and 3) trip (defined as duration between vehicle's ignition statuses ON to OFF). An algorithm for identification of ignition ON and OFF events is also presented, since this signal was not explicitly captured during the data acquisition phase. As a case study, these imputation techniques were applied to the data from a driver behavior classification experiment.