Self-driving Vehicles Research Articles

A deep reinforcement learning assisted task offloading and resource allocation approach towards self-driving object detection

With the development of communication technology and mobile edge computing (MEC), self-driving has received more and more research interests. However, most object detection tasks for self-driving vehicles are still performed at vehicle terminals, which often requires a trade-off between detection accuracy and speed. To achieve efficient object detection without sacrificing accuracy, we propose an end–edge collaboration object detection approach based on Deep Reinforcement Learning (DRL) with a task prioritization mechanism. We use a time utility function to measure the efficiency of object detection task and aim to provide an online approach to maximize the average sum of the time utilities in all slots. Since this is an NP-hard mixed-integer nonlinear programming (MINLP) problem, we propose an online approach for task offloading and resource allocation based on Deep Reinforcement learning and Piecewise Linearization (DRPL). A deep neural network (DNN) is implemented as a flexible solution for learning offloading strategies based on road traffic conditions and wireless network environment, which can significantly reduce computational complexity. In addition, to accelerate DRPL network convergence, DNN outputs are grouped by in-vehicle cameras to form offloading strategies via permutation. Numerical results show that the DRPL scheme is at least 10% more effective and superior in terms of time utility compared to several representative offloading schemes for various vehicle local computing resource scenarios.

Pyramidal 3D feature fusion on polar grids for fast and robust traversability analysis on CPU

Self-driving vehicles and autonomous ground robots require a reliable and accurate method to analyze the traversability of the surrounding environment for safe navigation. This paper proposes and evaluates a real-time machine learning-based traversability analysis method that combines geometric features with a pyramid-polar space representation based on SVM classifiers. In particular, we show that by fusing geometric features with information stemming from coarser pyramid levels that account for a broader space portion, as well as integrating important implementation details, allows for a noticeable boost in performance and reliability. The main goal of this work is to demonstrate that traversability analysis is possible with effective results and in real-time even on cheaper hardware than expensive GPUs, e.g. CPU-only PCs. The proposed approach has been compared with state-of-the-art deep learning approaches on publicly available datasets of outdoor driving scenarios, running such algorithms both on GPU and CPU to compare runtimes. Our method can be fully executed on CPU and achieves results close to the best-in-class methods, runs faster, and requires fewer and less expensive hardware resources, consuming less than 30% electrical power with respect to deep learning models on embedded processing units. We release with this paper the open-source implementation of our method.

B-AWARE: Blockage Aware RSU Scheduling for 5G Enabled Autonomous Vehicles

5G Millimeter Wave (mmWave) technology holds great promise for Connected Autonomous Vehicles (CAVs) due to its ability to achieve data rates in the Gbps range. However, mmWave suffers from a high beamforming overhead and requirement of line of sight (LOS) to maintain a strong connection. For Vehicle-to-Infrastructure (V2I) scenarios, where CAVs connect to roadside units (RSUs), these drawbacks become apparent. Because vehicles are dynamic, there is a large potential for link blockages. These blockages are detrimental to the connected applications running on the vehicle, such as cooperative perception and remote driver takeover. Existing RSU selection schemes base their decisions on signal strength and vehicle trajectory alone, which is not enough to prevent the blockage of links. Many modern CAVs motion planning algorithms routinely use other vehicle’s near-future path plans, either by explicit communication among vehicles, or by prediction. In this paper, we make use of the knowledge of other vehicle’s near future path plans to further improve the RSU association mechanism for CAVs. We solve the RSU association algorithm by converting it to a shortest path problem with the objective to maximize the total communication bandwidth. We evaluate our approach, titled B-AWARE, in simulation using Simulation of Urban Mobility (SUMO) and Digital twin for self-dRiving Intelligent VEhicles (DRIVE) on 12 highway and city street scenarios with varying traffic density and RSU placements. Simulations show B-AWARE results in a 1.05× improvement of the potential datarate in the average case and 1.28× in the best case vs. the state-of-the-art. But more impressively, B-AWARE reduces the time spent with no connection by 42% in the average case and 60% in the best case as compared to the state-of-the-art methods. This is a result of B-AWARE reducing nearly 100% of blockage occurrences.

Reverse Traceability Framework for Identifying Liability of Crashes for Self-Driving Vehicles Using Blockchains

Modern vehicles are increasingly having a higher level of technology and automation. Humans are increasingly becoming dependent on these modern technologies to take decisions related to their lives and safety. Such an increasing dependence on automation raises an important question. If an autonomous vehicle (AV) meets an accident, who will be responsible? It is not the human driver, but technology that makes those crucial decisions on the road. This question is attracting considerable attention in the insurance industry because traditional vehicle insurance is based on the liability of human drivers, but in the future, vehicle technology will replace human drivers. Therefore, the vehicle manufacturer or one of its suppliers may be held responsible for the accident. This paper presents a crash liability identification framework that can identify who is liable if there is a crash or an accident of an autonomous self-driving vehicle. The use cases demonstrate that the proposed framework can be used by regulators to efficiently identify the liable party when an AV crashes.

An efficient intelligent task management in autonomous vehicles using AIIOT and optimal kernel adaptive SVM

The fast evolution of artificial intelligence (AI) and the IoT gained more interest in the development of autonomous vehicles. The main challenges faced by autonomous car manufacturers are high computation costs and the lag of intelligent task management systems. The accident rates created by autonomous vehicles are increasing rapidly due to their unrestrained traffic, inaccurate location, and mapping methods. So, secure driving becomes the main concern in self-driving vehicle design. Moreover, the inadequate battery life and computation power made the system complex to minimize execution time as well as resource computation. Therefore, to handle all these complications, an intelligent task-managing system for autonomous vehicles is proposed in this paper. In this, each task is optimally executed by invoking the supervised resource predictor kernel data adaptive support vector machine-based multimodal bacterial foraging (KDASVM-MBF) method. The KDASVM-MBF intelligent task scheduling method is proposed to distribute all the tasks to the suitable processor based on central processing unit (CPU) usage and emergency. The proposed model is implemented in Python IDE-version 3.8 and examined using two multicore processors (Nvidia and AIIoT). The potential capability of the introduced type is evaluated by computing the performance methods such as response time, resource utilization, CPU utilization, execution time, prediction accuracy, and task miss rate. The experimental results reveal that the established KDASVM-MBF method accomplishes prediction accuracy of about 97% and 98% for Nvidia and AIIoT processors respectively with minimum task miss rate and execution time.

THE EVALUATION OF ROAD INFRASTRUCTURE FOR SELF-DRIVING VEHICLES

The article deals with the problems of the use of self-driving vehicles (SDV) on public the adaptation of road infrastructures well as with, the necessity and objectives of assessment, and the experience of foreign countries in this field. The main factors influencing for SDV were analyzed. An evaluation model of infrastructure adaptation for SDV has been developed, allowing rational planning and selection of the most appropriate selfdriving public transport routes. In order to efficiently integrate SDV into transport systems, it is recommended to evaluate the potential route alternatives using the SAW method and also to use the efficiency indicators proposed in the thesis and their significance. After analyzing the theoretical and practical aspects of road infrastructure adaptation for SDV conclusions and suggestions are given.

The Development of the Trust in Self-Driving Vehicles Scale (TSDV)

Self-driving vehicles (SDVs) are an emerging technology in which consumers have low levels of trust. Researchers/designers can understand and improve consumer trust through research and iterative design, but doing so effectively requires reliable measures. Although general trust-in-automation measures exist, a measure tailored to SDVs may provide a more accurate tool. This study presents work undertaken to create a domain specific trust measure for SDVs. Candidate items were given to 400 participants who rated their trust in an SDV portrayed in a narrative describing a ride in said vehicle. The Trust in Self-driving Vehicles Scale (TSDV) was created by analyzing participants’ responses using psychometric methods. Four factors were extracted from participants’ responses. Five items were retained for each factor to create the TSDV. Initial evidence of the validity of the instrument is presented through the TSDV’s ability to discriminate between a trustworthy and non-trustworthy vehicle, as portrayed in use scenarios.

Consumer Knowledge of and Confusion with Driving Automation: Trends, Challenges and Implications

A variety of lower-level automation features, including driver support technologies and active driving assist features, are widely implemented in today’s vehicles. However, consumer knowledge of driving automation remains limited, and a more comprehensive understanding is needed regarding the links between knowledge, acceptance, and consumer characteristics. This study used data from a multi-year series of large-scale surveys to analyze trends and dynamics in consumer knowledge. Results showed a consistently low accuracy regarding the definition of fully automated, self-driving vehicles across consumers of different demographic and behavioral characteristics. In contrast, self-rated knowledge of self-driving vehicles varied between different consumer segments, with no significant association with actual knowledge, suggesting a prevalence of misperceptions and over-confidence among the general public. This paper discusses implications for consistent messaging, targeted communication, and improved use of terminology for establishing knowledge and ensuring that future acceptance is grounded on correct understanding.

Indian traffic sign detection and recognition using deep learning

Traffic signs play a crucial role in managing traffic on the road, disciplining the drivers, thereby preventing injury, property damage, and fatalities. Traffic sign management with automatic detection and recognition is very much part of any Intelligent Transportation System (ITS). In this era of self-driving vehicles, calls for automatic detection and recognition of traffic signs cannot be overstated. This paper presents a deep-learning-based autonomous scheme for cognizance of traffic signs in India. The automatic traffic sign detection and recognition was conceived on a Convolutional Neural Network (CNN)- Refined Mask R-CNN (RM R-CNN)-based end-to-end learning. The proffered concept was appraised via an innovative dataset comprised of 6480 images that constituted 7056 instances of Indian traffic signs grouped into 87 categories. We present several refinements to the Mask R-CNN model both in architecture and data augmentation. We have considered highly challenging Indian traffic sign categories which are not yet reported in previous works. The dataset for training and testing of the proposed model is obtained by capturing images in real-time on Indian roads. The evaluation results indicate lower than 3% error. Furthermore, RM R-CNN’s performance was compared with the conventional deep neural network architectures such as Fast R-CNN and Mask R-CNN. Our proposed model achieved precision of 97.08% which is higher than precision obtained by Mask R-CNN and Faster R-CNN models.

Investigating Students' Attitudes & Trust in AI During COVID-19

Artificial Intelligence (AI) has been a part of society for a considerable amount of time since its formal foundation in 1956 and can be used to perform cognitive tasks at a level similar or greater than humans. In our modern era of advanced intelligence machines and processing capabilities, AI is being pushed further into society at a fast rate. With advancing AI like ChatGPT and self-driving vehicles, it is important for society to understand the implications of trusting this technology. Studies have discussed AI being used in the medical field to combat mental health issues and identify attitudes. However, there is a gap in research over adults' levels of trust towards AI and how those feelings have evolved over the course of COVID-19. This research paper explores how trust college/university undergraduate/graduate adults evolved over the COVID-19 pandemic and aims to fill that gap. To address the general attitudes and levels of trust towards AI, a mixed-method quantitative and qualitative study was conducted utilizing a survey. The survey inquired about general attitudes using the General Attitudes towards AI Scale (GAAIS) and open-ended questions to measure levels of trust in relation to the pandemic. It was concluded that while participants demonstrated a positive attitude towards AI (70.4% ), most participants had low levels of trust reflected in their fears and concerns of future AI implementation. A high percentage of positivity towards AI coupled with low levels of trust indicates a complex attitude towards artificial intelligence.

Localization for self-driving vehicles based on deep learning networks and RGB cameras

Localization for self-driving vehicles based on deep learning networks and RGB cameras

An intelligent transport system capable of collecting or foraging with many robotic vehicles: An intelligent computing paradigm

Self-driving vehicles are increasingly vital for various environmental tasks, such as data gathering, remote sensing, and mapping. They are particularly important in transportation, where advanced robotic vehicles can operate effectively without human oversight in challenging conditions. One exciting application is using collaborative autonomous robotic vehicles to locate specific points of interest like accidents, congestion, rain, fog, or icy roads. This article explores multi-robot systems for transportation tasks, considering interference that may impact the robot group’s performance. A mathematical model is analysed to quantify the influence of interference and transformed into a dimensionless form for analysis. The system is studied numerically, and a neural soft computing technique, LMA-NN, is proposed. The technique is established on a neural network. For performance evaluation of the LMA-NN technique, different performance benchmarks such as mean square error and absolute error with reference solution are employed. The results are also presented graphically.

A Driver Behavior Monitoring System for Sustainable Traffic and Road Construction

The perception and behavior of human drivers is of great importance for both sustainable road construction and sustainable coexistence with self-driving vehicles. To record the (hazard) perception and responsiveness of drivers, as many parameters as possible of the recognition and reaction process must be measured and recorded. It must be known what the drivers see or perceive and how they react to it. In modern vehicles, all pedal positions and driving data are transmitted to the CAN bus and can be assigned to the associated sources using reverse engineering methods. What drivers see or recognize can be recorded from the driver’s perspective using (eye-tracking) videos or photos. To evaluate the vehicle and image data, systems are required that can, among other things, record both channels and link them synchronously. To fulfill this purpose, two different systems are developed and compared here, both of which can be used in real traffic.

Research on Lane-Change Decision and Planning in Multilane Expressway Scenarios for Autonomous Vehicles

Taking into account the issues faced by self-driving vehicles in multilane expressway scenarios, a lane-change decision planning framework that considers two adjacent lanes is proposed. Based on this framework, the lateral stability of an autonomous vehicle under near-limit conditions during lane change is studied by the phase-plane method. Firstly, a state-machine-based driving logic is designed and a decision method is proposed to design the lane-change intention based on the surrounding traffic information and to consider the influence of the motion state of other vehicles in the adjacent lanes on the self-driving vehicle. In order to realize adaptive cruising under the full working conditions of the vehicle, a safety distance model is established for different driving speeds and switching strategies for fixed-speed cruising, following driving, and emergency braking are developed. Secondly, for the trajectory planning problem, a lane-change trajectory based on a quintuple polynomial optimization method is proposed. Then, the vehicle lateral stability boundary is investigated; the stability boundary and rollover boundary are incorporated into the designed path-tracking controller to improve the tracking accuracy while enhancing the rollover prevention capability. Finally, a simulation analysis is carried out through a joint simulation platform; the simulation results show that the proposed method can ensure the driving safety of autonomous vehicles in a multilane scenario.

An Improved Lightweight Network Using Attentive Feature Aggregation for Object Detection in Autonomous Driving

Object detection, a more advanced application of computer vision than image classification, utilizes deep neural networks to predict objects in an input image and determine their locations through bounding boxes. The field of artificial intelligence has increasingly focused on the demands of autonomous driving, which require both high accuracy and fast inference speeds. This research paper aims to address this demand by introducing an efficient lightweight network for object detection specifically designed for self-driving vehicles. The proposed network, named MobDet3, incorporates a modified MobileNetV3 as its backbone, leveraging its lightweight convolutional neural network algorithm to extract and aggregate image features. Furthermore, the network integrates altered techniques in computer vision and adjusts to the most recent iteration of the PyTorch framework. The MobDet3 network enhances not only object positioning ability but also the reusability of feature maps across different scales. Extensive evaluations were conducted to assess the effectiveness of the proposed network, utilizing an autonomous driving dataset, as well as large-scale everyday human and object datasets. These evaluations were performed on NXP BlueBox 2.0, an advanced edge development platform designed for autonomous vehicles. The results demonstrate that the proposed lightweight object detection network achieves a mean precision of up to 58.30% on the BDD100K dataset and a high inference speed of up to 88.92 frames per second on NXP BlueBox 2.0, making it well-suited for real-time object detection in autonomous driving applications.

Multiple UAVs collaborative traffic monitoring with intention-based communication

This article proposes an innovative ground traffic management approach that utilizes the mobility, flexibility, and collaborative capabilities of multiple Unmanned Aerial Vehicles (UAVs). The objective is to enhance the navigation and driving experience of self-driving vehicles by employing UAVs to cover blind areas that cannot be observed by ground monitoring equipment, thereby facilitating the avoidance of congested routes. This study focuses on determining the optimal number of UAVs and enhancing the scheduling strategy and task assignment method by introducing novel UAV communication collaboration techniques. The research faces challenges such as a limited quantity of UAVs, rapid response and feedback, variations in traffic conditions and tasks, and system complexity. Furthermore, the difficulty of communication collaboration is exacerbated by the incomplete connectivity of UAV networks. To address these challenges, this article introduces a method that utilizes intention information as the message for communication between UAVs, thereby enhancing their collaborative capabilities. Using the Multi-Agent Reinforcement Learning (MARL) neural network, this method generates intention information and utilizes the intention information generated by other UAVs to optimize the flight control and traffic monitoring task allocation strategy. This method effectively addresses the challenge of UAV collaboration in efficiently managing a large volume of stochastic traffic monitoring tasks. Experimental results demonstrate that the proposed approach significantly improves vehicle navigation accuracy and achieves a more balanced distribution of workload among multiple UAVs. Moreover, by sharing road conditions detected in the cloud, it reduces operational costs and opens up possibilities for the practical implementation of UAVs in traffic management.

SWOT Analysis on AI-based Self-driving Car Companies

Purpose: This study examines the challenges and possibilities that advancing self-driving vehicle technologies may provide for transportation regulations. Self-driving technology can benefit low-income households and people with mobility issues since it can increase accessibility while reducing dependability and commuting expenses. This emerging technology has uses and effects that go far beyond any current projections. This study offers a complete review of the relevant papers and literature, covering a range of topics from security to viability from an economic perspective. In this regard, we have analyzed certain parameters on the challenges faced for launching self-driving vehicles. According to the results of the SWOT analysis, a step-by-step strategy to introduce self-driving vehicles is the right approach. The export of self-driving vehicles can help to create jobs and maintain economic stability. Methodology: This study is based on exploratory research by examining various reports from websites and recent research papers. Further, we have accessed and framed Strength-Opportunity (SO), Weakness-Opportunity (WO), Strength-Threat (ST), and Weakness- Threat (WT) strategies for the smooth implementation of self-driving vehicles in India. Finding: The present study will analyze the opportunity in implementing self driving vehicle and factors like exports that will influence Indian economy. Government is rigid with self driving cars because of the fact that many will lose their jobs. Originality: This study addresses a number of concerns, prominent among them is the difficulty in adopting self-driving cars and their effects once they become operational or are made legal. Paper Type: Exploratory based research analysis on challenges for implementation of self driving vehicle.

Obtaining advantages from knowledge base: mapping the potential to develop new technologies

ABSTRACT It is necessary for the company to find suitable new growth points when growth hits the ceiling. Technology opportunity discovery is one kind of knowledge based risk mitigation strategy (TOD). Computer-based TOD has the advantage of fast and repeatable analysis, which can be used as a complement and reference to expert decision methods. This research shows an approach to help firms to discover technology opportunities based on their knowledge foundation to obtain a competitive advantage. The proposed method involves (1) identifying the critical technologies by text analysis, measuring the importance of technology in the network and the time of bursts, (2) evaluating the relative technological advantage (RTA) of the firms as knowledge base, (3) Calculating the relatedness between technologies, assessing the development potential of non-RTA technologies from the firm’s current knowledge base and the relatedness between technologies, (4) proposing a technology opportunity analysis framework that integrates the potential and importance for technological development. This article chooses Self-Driving Vehicles (SDV) as a case study to further explain and validate the approach. This approach helps firms understand their current state and growth potential. Such recommendations are based on the knowledge foundation of different firms and therefore are individualised.

An Efficient Safety-Oriented Car-Following Model for Connected Automated Vehicles Considering Discrete Signals

With the rapid development of Connected and Automated Vehicle (CAV) technology, limited self-driving vehicles have been commercially available in certain leading intelligent transportation system countries. When formulating the car-following model for CAVs, safety is usually the basic constraint. Safety-oriented car-following models seek to specify a safe following distance that can guarantee safety if the preceding vehicle were to brake hard suddenly. The discrete signals of CAVs bring a series of phenomena, including discrete decision-making, phase difference, and discretely distributed communication delay. The influences of these phenomena on the car-following safety of CAVs are rarely considered in the literature. This paper proposes an efficient safety-oriented car-following model for CAVs considering the impact of discrete signals. The safety constraints during both normal driving and a sudden hard brake are incorporated into one integrated model to eliminate possible collisions during the whole driving process. The mechanical delay information of the preceding vehicle is used to improve car-following efficiency. Four modules are designed to enhance driving comfort and string stability in case of heavy packet losses. Simulations of a platoon with diversified vehicle types demonstrate the safety, efficiency, and string stability of the proposed model. Tests with different packet loss rates imply that the model could guarantee safety and driving comfort in even poor communication environments.

Scene Detection Classification and Tracking for Self-Driven Vehicle

A number of traffic-related issues, including crashes, jams, and pollution, could be resolved by self-driving vehicles (SDVs). Several challenges still need to be overcome, particularly in the areas of precise environmental perception, observed detection, and its classification, to allow the safe navigation of autonomous vehicles (AVs) in crowded urban situations. This article offers a comprehensive examination of the application of deep learning techniques in self-driving cars for scene perception and observed detection. The theoretical foundations of self-driving cars are examined in depth in this research using a deep learning methodology. It explores the current applications of deep learning in this area and provides critical evaluations of their efficacy. This essay begins with an introduction to the ideas of computer vision, deep learning, and self-driving automobiles. It also gives a brief review of artificial general intelligence, highlighting its applicability to the subject at hand. The paper then concentrates on categorising current, robust deep learning libraries and considers their critical contribution to the development of deep learning techniques. The dataset used as label for scene detection for self-driven vehicle. The discussion of several strategies that explicitly handle picture perception issues faced in real-time driving scenarios takes up a sizeable amount of the work. These methods include methods for item detection, recognition, and scene comprehension. In this study, self-driving automobile implementations and tests are critically assessed.