Deployment Of Autonomous Vehicles Research Articles

Advancements in Machine Learning Techniques for Traffic Flow Prediction in Autonomous Vehicles: A Comprehensive Review

Traffic flow prediction is a critical aspect of enabling the successful deployment of autonomous vehicles (AVs) in urban environments. Accurate and reliable traffic flow prediction plays a crucial role in allowing AVs to navigate efficiently and make informed decisions. Machine learning techniques have emerged as powerful tools for traffic flow prediction, leveraging large-scale datasets and complex models to capture the inherent dynamics of traffic patterns. This comprehensive review examines the state-of-the-art machine learning techniques used for traffic flow prediction in AVs, highlighting their strengths, limitations, and future research directions. The review also explores data sources and preprocessing techniques, performance evaluation metrics, case studies, and real-world applications. Furthermore, it discusses the challenges associated with traffic flow prediction for AVs, such as data scarcity and model interpretability, and identifies promising future research directions, including reinforcement learning and multimodal fusion. This review serves as a valuable resource for researchers, practitioners, and policymakers interested in advancing traffic flow prediction for autonomous vehicles

Overweight and overload implementation in Apollo systems

Apollo is a high-performance, flexible architecture developed by Baidu, Kinglong, and a consortium of more than 40 companies for the purpose of accelerating the development, testing, and deployment of Autonomous Vehicles. This article first describes Apollo in terms of the functionality of the system and the interaction between its components, and then analyzes the conceptual architecture through system evolution, control and data flow, and concurrency, while illustrating the impact of the division of developer responsibilities on this. Then will go through a detailed overview of the concrete architecture of Baidu Apollo with working on mapping the source code from the Apollo Github Website to our architecture using the diagram drawing tool Scitools Understand. And divided into five parts, the report concludes with five major parts: the process from mapping code to diagram; concrete architecture explanation; unexpected dependencies discovery; subsystem analysis, and sequence diagrams presented in a more accurate and specific way. Finally, this report will propose a specific feature or enhancement to the current concrete architecture. The impact that this feature or enhancement can have on the non-functional requirements and stakeholders of the system will be explored based on the existing architecture and component interactions. In this regard, this report discusses two specific approaches that can be implemented and compares the two approaches through SAAM analysis to determine the best approach.

Investigating how the public acceptance of autonomous vehicles evolve with the changes in the level of knowledge: A demographic analysis

Autonomous vehicles (AVs) are expected to provide various advantages such as improved mobility, increased comfort, and reduced traffic accidents. However, the deployment of AVs is contingent upon the public’s attitude, which in turn may affect their consequences. The relationship between public knowledge about AVs and attitude has been debated. While some studies indicate a positive correlation between knowledge and optimism, others suggest a negative association. The present study aims to evaluate the association between public knowledge, attitude, and AVs in the US. A questionnaire survey was conducted between June and November 2022, yielding 5778 complete responses from various regions of the US. Data were analyzed to estimate the public attitude and level of knowledge by region. Findings revealed a negative shift in public attitude towards AVs with increased knowledge. Specifically, a 1% increase in knowledge was associated with a 0.65% reduction in interest, a 0.68% decrease in trust, and a decline of $2466 USD in willingness to pay for AVs, as well as a 0.56% increase in concerns about traveling in AVs. In addition, further analysis was conducted to understand how the public attitude of different demographic groups evolves with the level of knowledge. Furthermore, the results of this study were discussed in light of the introduction of the automobiles showing a lot of similarities in the public attitude towards the introduction of both the automobiles (more than 100 years ago) and AVs (now). In addition, the results were discussed in light of the theory of diffusion of innovation and the Gartner Hype Cycle which are used to explain how the public reacts to innovations.

Online Adaptive Generation of Critical Boundary Scenarios for Evaluation of Autonomous Vehicles

Testing and evaluation are critical steps in the development and deployment of autonomous vehicles (AVs). This paper aims to provide an online adaptive generation framework for critical boundary driving scenarios (CBDS) with flexible complexity and diversity. Both static scenarios and dynamic scenarios are comprehensively investigated based on real traffic flow data, where the generated static scenarios contain variables of drivable area, weather visibility, and road friction coefficient on autonomous vehicles, and the generated dynamic scenarios consider the effects of mixed traffic streams of AVs, human-driven vehicles, bicycles, and pedestrians. Two complexity models are proposed to characterize the complexity of static scenarios and dynamic scenarios separately, which help to construct feedback for the adaptive generation of CBDS. The scenario-based test is carried out on a joint simulation platform, and the multi-dimensional evaluation system, including safety, driving comfort, driving performance, and traffic coordination, is developed to assess the performance of AVs. Based on the proposed complexity models and multi-dimensional evaluation system, the generation of CBDS is transformed into online environmental parameter optimization. Moreover, the naturalistic driving scenarios generation, scenario complexity calculation, intelligent driving algorithm execution, and intelligent driving evaluation are all concatenated together to form a closed loop so as to adaptively generate critical boundary driving scenarios. Extensive simulations are conducted at the intersection with two different types of intelligent driving algorithms, which show the effectiveness of the proposed framework.

Robust real-time traffic light detector on small-form platform for autonomous vehicles

Timely and accurate detection and recognition of traffic lights are critical for Autonomous Vehicles (AVs) to avoid crashes due to red light running. This paper integrates a new robust machine learning based solution by combining a Convolutional Neural Network (CNN) with computer vision techniques to achieve a real-time traffic light detector. The proposed detection and recognition algorithm is capable of recognizing traffic lights on low-power small-form platforms, which are lightweight, portable, and can be mounted on AVs in daylight scenarios. The LISA open-source dataset is utilized with augmentation methods to increase the accuracy of the solution. The proposed approach achieves 93.42% of accuracy at a speed of 30.01 Frames Per Second (FPS) on an NVIDIA Jetson Xavier platform without using hardware accelerators such as FPGA. This solution is expected to promote the quicker adoption and wider deployment of AVs by increasing the chances of avoiding crashes and ultimately saving lives.

Risk-aware controller for autonomous vehicles using model-based collision prediction and reinforcement learning

Autonomous Vehicles (AVs) have the potential to save millions of lives and increase the efficiency of transportation services. However, the successful deployment of AVs requires tackling multiple challenges related to modeling and certifying safety. State-of-the-art decision-making methods usually rely on end-to-end learning or imitation learning approaches, which still pose significant safety risks. Hence the necessity of risk-aware AVs that can better predict and handle dangerous situations. Furthermore, current approaches tend to lack explainability due to their reliance on end-to-end Deep Learning, where significant causal relationships are not guaranteed to be learned from data.This paper introduces a novel risk-aware framework for training AV agents using a bespoke collision prediction model and Reinforcement Learning (RL). The collision prediction model is based on Gaussian Processes and vehicle dynamics, and is used to generate the RL state vector. Using an explicit risk model increases the post-hoc explainability of the AV agent, which is vital for reaching and certifying the high safety levels required for AVs and other safety-sensitive applications. Experimental results obtained with a simulator and state-of-the-art RL algorithms show that the risk-aware RL framework decreases average collision rates by 15%, makes AVs more robust to sudden harsh braking situations, and achieves better performance in both safety and speed when compared to a standard rule-based method (the Intelligent Driver Model). Moreover, the proposed collision prediction model outperforms other models in the literature.

Legal Framework for Rear-End Crashes in Mixed-Traffic Platooning: A Matrix Game Approach

Autonomous vehicles (AV) hold great potential to increase road safety, reduce traffic congestion, and improve mobility systems. However, the deployment of AVs introduces new liability challenges when they are involved in car accidents. A new legal framework should be developed to tackle such a challenge. This paper proposes a legal framework, incorporating liability rules to rear-end crashes in mixed-traffic platoons with AVs and human-propelled vehicles (HV). We leverage a matrix game approach to understand interactions among players whose utility captures crash loss for drivers according to liability rules. We investigate how liability rules may impact the game equilibrium between vehicles and whether human drivers’ moral hazards arise if liability is not designed properly. We find that compared to the no-fault liability rule, contributory and comparative rules make road users have incentives to execute a smaller reaction time to improve road safety. There exists moral hazards for human drivers when risk-averse AV players are in the car platoon.

Dense reinforcement learning for safety validation of autonomous vehicles.

One critical bottleneck that impedes the development and deployment of autonomous vehicles is the prohibitively high economic and time costs required to validate their safety in a naturalistic driving environment, owing to the rarity of safety-critical events1. Here we report the development of an intelligent testing environment, where artificial-intelligence-based background agents are trained to validate the safety performances of autonomous vehicles in an accelerated mode, without loss of unbiasedness. From naturalistic driving data, the background agents learn what adversarial manoeuvre to executethrough a dense deep-reinforcement-learning (D2RL) approach, in which Markov decision processes are edited by removing non-safety-critical states and reconnecting critical ones so that the information in the training data is densified. D2RL enables neural networks to learn from densified information with safety-critical events and achieves tasks that are intractable for traditional deep-reinforcement-learning approaches. We demonstrate the effectiveness of our approach by testing a highly automated vehicle in both highway and urban test tracks with an augmented-reality environment, combining simulated background vehicles with physical road infrastructure anda real autonomous test vehicle. Our results show that the D2RL-trained agents can accelerate the evaluation process by multiple orders of magnitude (103 to 105 times faster). In addition, D2RL will enable accelerated testing and training with other safety-critical autonomous systems.

Quantifying the Impact of Deployments of Autonomous Vehicles and Intelligent Roads on Road Safety in China: A Country-Level Modeling Study.

Approximately 1.35 million people lose their lives due to road traffic collisions worldwide per year. However, the variation of road safety depending on the deployment of Autonomous Vehicles (AV), Intelligent Roads (IR), and Vehicle-to-Vehicle technology (V2V) is largely unknown. In this analysis, a bottom-up analytical framework was developed to evaluate the safety benefits of avoiding road injuries and reducing crash-related economic costs from the deployment of AVs, IRs, and V2Vs in China in 26 deployment scenarios from 2020 to 2050. The results indicate that compared with only deploying AVs, increasing the availability of IRs and V2V while reducing the deployment of fully AVs can achieve larger safety benefits in China. Increasing the deployment of V2V while reducing the deployment of IRs can sometimes achieve similar safety benefits. The deployment of AVs, IRs, and V2V plays different roles in achieving safety benefits. The large-scale deployment of AVs is the foundation of reducing traffic collisions; the construction of IRs would determine the upper limit of reducing traffic collisions, and the readiness of connected vehicles would influence the pace of reducing traffic collisions, which should be designed in a coordinated manner. Only six synergetic scenarios with full equipment of V2V can meet the SDG 3.6 target for reducing casualties by 50% in 2030 compared to 2020. In general, our results highlight the importance and the potential of the deployment of AVs, IRs, and V2V to reduce road fatalities and injuries. To achieve greater and faster safety benefits, the government should prioritize to the deployment of IRs and V2V. The framework developed in this study can provide practical support for decision-makers to design strategies and policies on the deployment of AVs and IRs, which can also be applied in other countries.

Opportunities and Directions for the Evolution of Command and Control Systems in the Context of Multi-domain Operations

The military conflicts of recent times have highlighted the need for modern military operations to focus on multi-domain operations, which requires a transformation of the command and control system, by creating mobile and flexible headquarters that can provide relevant information at any time, even during redeployments and redeployments. Cloud computing can be used to integrate data collection tools, ensuring fast and accurate analysis of large amounts of incoming data. A private 5G network can provide a secure, high-speed and reliable communication environment for cloud-based command and control systems. This network will enable real-time information transfer, facilitate rapid decision-making processes and enable the deployment of autonomous vehicles and drones. The authors have also explored capabilities that support the deployment of autonomous vehicles and drones, enabling commanders to conduct intelligence and surveillance more effectively. By leveraging technology and advanced communications systems, commanders can lead their teams in a dynamic environment while providing a high level of situational awareness. This integration of information and mobility allows commanders to react quickly and decisively, ultimately increasing their effectiveness on the battlefield.

Effect of Cyber Vulnerabilities on the Adoption of Self-Driving Vehicles – A Review

One of the leading disruptive technologies in the upcoming technological revolution is Self-Driving vehicles. However, the absence of security is the greatest obstacle to adoption. This study looks at how cybersecurity impacts the adoption of driverless cars. The purpose of this paper is to perform a literature review supporting the in-depth analysis of cybersecurity and its impacts on the slower adoption rate of Self-Driving Vehicles. The study's primary goal is to determine the connection between worries about cybersecurity and the rate of adoption of self-driving vehicles. Driverless vehicles are the most effective and cutting-edge technology in the transportation sector, yet there are barriers to their widespread adoption because of cybersecurity worries. As a result, this study will clarify the cybersecurity issues that contributed to the slower deployment of autonomous vehicles. The NIST Cybersecurity Framework serves as the study's theoretical foundation. This paradigm consistently identifies the barriers to new technology adoption in cybersecurity.

Sustainable Deployment of Autonomous Vehicles Dedicated Lanes in Urban Traffic Networks

Sustainable Deployment of Autonomous Vehicles Dedicated Lanes in Urban Traffic Networks

AdaSplats: Adaptive Splatting of Point Clouds for Accurate 3D Modeling and Real-Time High-Fidelity LiDAR Simulation

LiDAR sensors provide rich 3D information about their surroundings and are becoming increasingly important for autonomous vehicles tasks such as localization, semantic segmentation, object detection, and tracking. Simulation accelerates the testing, validation, and deployment of autonomous vehicles while also reducing cost and eliminating the risks of testing in real-world scenarios. We address the problem of high-fidelity LiDAR simulation and present a pipeline that leverages real-world point clouds acquired by mobile mapping systems. Point-based geometry representations, more specifically splats (2D oriented disks with normals), have proven their ability to accurately model the underlying surface in large point clouds, mainly with uniform density. We introduce an adaptive splat generation method that accurately models the underlying 3D geometry to handle real-world point clouds with variable densities, especially for thin structures. Moreover, we introduce a fast LiDAR sensor simulator, working in the splatted model, that leverages the GPU parallel architecture with an acceleration structure while focusing on efficiently handling large point clouds. We test our LiDAR simulation in real-world conditions, showing qualitative and quantitative results compared to basic splatting and meshing techniques, demonstrating the interest of our modeling technique.

Modeling Autonomous Vehicles Deployment in a Multilane AV Zone With Mixed Traffic

The advancement of technologies has fast-tracked the deployment of autonomous vehicles (AVs), however, they necessitate the modernization of the existing road infrastructure to meet the technical and safety requirements of AVs. Due to the high cost and prolonged time required for upgrading infrastructure, it is expected that AVs will be introduced gradually into the existing transport network of human-driven vehicles (HVs) either jointly or separately in a special area or zone within that network. In this paper, we propose a mathematical framework for modelling and deployment of multi-lane AV zone in a mixed-user traffic network that takes into account the different route choice decisions of HVs and AVs. The model is used to study two different designs for a multi-lane AV zone, where in Design 1, both lanes are dedicated for AVs inside the AV zone; while in Design 2, one lane is dedicated for AVs, and another lane is for the HV traffic. We then derive an equivalent convex formulation and develop a novel solution method for this complex problem, as well as evaluate and compare the performance of the designs in a realistic large-scale transportation network. Numerical results show that the deployment of a multi-lane AV zone improves network performance in terms of the total system travel time (TSTT). The proposed framework can be used to assist transportation planners and decision-makers to successfully deploy AVs seamlessly with mixed traffic.

Disengagement Cause-and-Effect Relationships Extraction Using an NLP Pipeline

The advancement in machine learning and artificial intelligence promotes the testing and deployment of autonomous vehicles (AVs) on public roads. The California Department of Motor Vehicles (CA DMV) has launched the Autonomous Vehicle Tester Program, which collects and releases reports related to Autonomous Vehicle Disengagement (AVD) from autonomous driving. Understanding the causes of AVD is critical to improving the AV system’s safety and stability and providing guidance for AV testing and deployment. In this work, we built a scalable end-to-end pipeline to collect, process, model, and analyze the disengagement reports released from 2014 to 2020 using natural language processing and deep transfer learning. The analysis of disengagement data using taxonomy, visualization, and statistical tests revealed the trends of AV testing, cause frequency, and significant relationships between causes and effects of AVD. We found that (1) manufacturers tested AVs intensively during the Spring and/or Winter, (2) test drivers initiated more than 80% of the disengagement while more than 75% of the disengagement were because of errors in perception, localization & mapping, planning and control of the AV system, and (3) there was a significant relationship between the initiator of AVD and the cause category. This study serves as a successful practice of deep transfer learning using pre-trained models and generates a consolidated disengagement database allowing further investigation for other researchers. The related code and data are available on github. <xref ref-type="fn" rid="fn1" xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</xref>

On the Role of Futuristic Technologies in Securing UAV-Supported Autonomous Vehicles

Unmanned aerial vehicles (UAVs) and autonomous vehicle (AV) technologies are gaining popularity due to their key benefits that include, but are not limited to, infrastructure independence, scalability, and autonomy. Although UAV and AV technologies are vertical; nevertheless, from the services and applications standpoint, these technologies handsomely complement each other. More precisely, the shortcomings faced by AV technology such as real-time navigation, visibility, mapping, content uploading, and maneuvering could be supported by UAV technology. Moreover, security plays a crucial role in the commercial deployment of AVs. This article focuses on the role of futuristic technologies in securing such a critical environment. Specifically, we focus on the data-related security for which blockchain is leveraged. The role of other enabling technologies, such as artificial intelligence, in providing security to the integrated UAV–AV has also been discussed. Finally, we identify the outstanding research and development challenges that will provide new research avenues for the research community.

E-Scooter Rider detection and classification in dense urban environments

Accurate detection and classification of vulnerable road users is a safety critical requirement for the deployment of autonomous vehicles in heterogeneous traffic. Although similar in physical appearance to pedestrians, e-scooter riders follow distinctly different characteristics of movement and can reach speeds of up to 45 kmph. The challenge of detecting e-scooter riders is exacerbated in urban environments where the frequency of partial occlusion is increased as riders navigate between vehicles, traffic infrastructure and other road users. This can lead to the non-detection or mis-classification of e-scooter riders as pedestrians, providing inaccurate information for accident mitigation and path planning in autonomous vehicle applications. This research introduces a novel benchmark for partially occluded e-scooter rider detection to facilitate the objective characterization of detection models. A novel, occlusion-aware method of e-scooter rider detection is presented that achieves a 15.93% improvement in detection performance over the current state of the art.

GNSS-Imaging Data Fusion for Integrity Enhancement in Autonomous Vehicles

The transport sector is experiencing a fast and revolutionary development, moving toward the deployment of autonomous vehicles. Thus, the accuracy of the position information and the integrity of the navigation system have become key factors. In this article, we address the problem of the enhancement of the integrity of the position provided by a global navigation satellite system receiver by exploiting sensor fusion. To this aim, we estimate the lateral offset and heading of the vehicle with respect to a georeferenced roadway centerline from the images supplied by an on-board camera. Moreover, we perform integrity monitoring based on the implementation of the solution separation in the parity space. The numerical results indicate that the use of sensor fusion and digital map allows us to attain a longitudinal protection level reduction with respect to the case in which sensor fusion is not exploited. More specifically, a decrease of about 70% is achieved when a single constellation is used, while reduction is less relevant, about 15%, when two constellations are employed.

Improving Worst Case Visual Localization Coverage via Place-Specific Sub-Selection in Multi-Camera Systems

6-DoF visual localization systems utilize principled approaches rooted in 3D geometry to perform accurate camera pose estimation of images to a map. Current techniques use hierarchical pipelines and learned 2D feature extractors to improve scalability and increase performance. However, despite gains in typical recall <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$@0.25\,{\rm{m}}$</tex-math></inline-formula> type metrics, these systems still have limited utility for real-world applications like autonomous vehicles because of their <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">worst</i> areas of performance - the locations where they provide insufficient recall at a certain required error tolerance. Here we investigate the utility of using <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">place specific configurations</i> , where a map is segmented into a number of places, each with its own configuration for modulating the pose estimation step, in this case selecting a camera within a multi-camera system. On the Ford AV benchmark dataset, we demonstrate substantially improved worst-case localization performance compared to using off-the-shelf pipelines - minimizing the percentage of the dataset which has low recall at a certain error tolerance, as well as improved overall localization performance. Our proposed approach is particularly applicable to the crowdsharing model of autonomous vehicle deployment, where a fleet of AVs are regularly traversing a known route.

Extensible Co-Simulation Framework for Supporting Cooperative Driving Automation Research

Autonomous vehicles (AVs) and cooperative automated vehicles (CAVs) are expected to largely reshape our mobility systems. The limited deployment of AVs and CAVs on roads makes it difficult to fully assess their impact and interactions with other road users. Advanced simulations are often sought for conducting accelerated tests of AVs and CAVs in a virtual environment. However, existing off-the-shelf simulators are typically focused on conventional traffic simulation and human-driving simulation. Advanced simulators that enable core functionalities (e.g., sensing and communication) of AVs and CAVs have been underexploited. In this paper, the authors aim to develop a realistic co-simulation framework for testing autonomous driving and cooperative driving automation (CDA). The proposed co-simulation framework utilizes the open-source concept to support the AV and CAV community in developing and deploying AV and CAV technologies. This framework integrates multiple open-source platforms, including Eclipse MOSAIC™ simulation framework, Eclipse Simulation of Urban Mobility (SUMO™) traffic simulator, and CARLA AV driving simulator. The framework enables AV and CAV simulation in mixed traffic environments. The developed co-simulation models have been tested with different scales of networks and traffic flow. The assessment of the co-simulation framework shows that it can support faster-than-real-time simulation for use in accelerated tests with more realistic scenarios. In addition, the developed co-simulation framework is proven to be extensible with the inclusion of other network simulators for supporting vehicle-to-everything (V2X) communication among vehicles.