Search Results

Roundabouts are one of the most complex traffic environments in urban roads, and a key challenge for intelligent driving decision-making. Deep reinforcement learning, as an emerging solution for intelligent driving decisions, has the advantage of avoiding complex algorithm design and sustainable iteration. For the decision difficulty in roundabout scenarios, this paper proposes an artificial potential field based Soft Actor-Critic (APF-SAC) algorithm. Firstly, based on the Carla simulator and Gym framework, a reinforcement learning simulation system for roundabout driving is built. Secondly, to reduce reinforcement learning exploration difficulty, global path planning and path smoothing algorithms are designed to generate and optimize the path to guide the agent.

The European Commission is going to publish the new Euro7 standard shortly, with the target of reducing the impact on pollutant emissions due to transportation systems. Besides forcing internal combustion engines to operate cleaner in a wider range of operating conditions, the incoming regulation will point out the role of On-Board Monitoring (OBM) as a key enabler to ensure limited emissions over the whole vehicle lifetime, necessarily taking into account the natural aging of involved systems and possible electronic/mechanical faults and malfunctions. In this scenario, this work aims to study the potential of data-driven approaches in detecting emission-relevant engine faults, supporting standard On-Board Diagnostics (OBD) in pinpointing faulty components, which is part of the main challenges introduced by Euro7 OBM requirements.

For intelligent vehicles, a robust perception system relies on training datasets with a large variety of scenes. The architecture of federated learning allows for efficient collaborative model iteration while ensuring privacy and security by leveraging data from multiple parties. However, the local data from different participants is often not independent and identically distributed, significantly affecting the training effectiveness of autonomous driving perception models in the context of federated learning. Unlike the well-studied issues of label distribution discrepancies in previous work, we focus on the challenges posed by scene heterogeneity in the context of federated learning for intelligent vehicles and the inadequacy of a single scene for training multi-task perception models. In this paper, we propose a federated learning-based perception model training system.

With economic development and the increasing number of vehicles in cities, urban transport systems have become an important issue in urban development. Efficient traffic signal control is a key part of achieving intelligent transport. Reinforcement learning methods show great potential in solving complex traffic signal control problems with multidimensional states and actions. Most of the existing work has applied reinforcement learning algorithms to intelligently control traffic signals. In this paper, we investigate the agent-based reinforcement learning approach for the intelligent control of ramp entrances and exits of urban arterial roads, and propose the Proximal Policy Optimization (PPO) algorithm for traffic signal control. We compare the method controlled by the improved PPO algorithm with the no-control method.

The prediction of agents' future trajectory is a crucial task in supporting advanced driver-assistance systems (ADAS) and plays a vital role in ensuring safe decisions for autonomous driving (AD). Currently, prevailing trajectory prediction methods heavily rely on high-definition maps (HD maps) as a source of prior knowledge. While HD maps enhance the accuracy of trajectory prediction by providing information about the surrounding environment, their widespread use is limited due to their high cost and legal restrictions. Furthermore, due to object occlusion, limited field of view, and other factors, the historical trajectory of the target agent is often incomplete This limitation significantly reduces the accuracy of trajectory prediction. Therefore, this paper proposes ETSA-Pred, a mapless trajectory prediction model that incorporates enhanced temporal modeling and spatial self-attention.

Fracture characterization of automotive metals under simple shear deformation is critical for the calibration of advanced fracture models employed in forming and crash simulations. In-plane shear fracture tests of high ductility materials have proved challenging since the sample edge fails first in uniaxial tension before the fracture limit in shear is reached at the center of the gage region. Although through-thickness machining is undesirable, it appears required to promote higher strains within the shear zone. The present study seeks to adapt existing in-plane shear geometries, which have otherwise been successful for many automotive materials, to have a local shear zone with a reduced thickness. It is demonstrated that a novel shear zone with a pocket resembling a “peanut” can promote shear fracture within the shear zone while reducing the risk for edge fracture. An emphasis was placed upon machinability and surface quality for the design of the pocket in the shear zone.

It is very important to secure the purity of the sound source to improve the degree of development of the noise problem, which is one of the important factors in vehicle development. So far, to acquire only the noise of the component, which is a problem element in vehicle driving noise, the component is removed and driven to acquire the noise, or the method of denoising the noise of other parts has been used. However, the method of removing part takes a lot of time to remove the part, and when the noise of the removed part is acquired, it has a disadvantage in that it differs from the characteristics of the noise measured in the mounting state of the vehicle. In addition, the method of denoising may cause data loss due to the deformation of the sound source of the noise.

Multiple hybrid bead designs were investigated in this study to control the springback on DP780 samples using post-stretching technique. The performance of the four different hybrid bead designs was evaluated by measuring the minimum blank-lock tonnage required to control the springback during a U-channel stamping process. A finite element (FE) model of the U-channel stamping process was developed to simulate the process and predict the minimum blank-lock tonnage required for springback control using each of the hybrid bead designs. It is shown that the developed FE model predicts both the required minimum blank-lock tonnage for post-stretching, and the springback profile, with good accuracy.

This paper explores the role and challenges of Artificial Intelligence (AI) algorithms, specifically AI-based software elements, in autonomous driving systems. These AI systems are fundamental in executing real-time critical functions in complex and high-dimensional environments. They handle vital tasks like multi-modal perception, cognition, and decision-making tasks such as motion planning, lane keeping, and emergency braking. A primary concern relates to the ability (and necessity) of AI models to generalize beyond their initial training data. This generalization issue becomes evident in real-time scenarios, where models frequently encounter inputs not represented in their training or validation data. In such cases, AI systems must still function effectively despite facing distributional or domain shifts. This paper investigates the risk associated with overconfident AI models in safety-critical applications like autonomous driving.

During the vehicle lifecycle, customers are able to directly perceive the outer panel stiffness of vehicles in various environmental conditions. The outer panel stiffness is an important factor for customers to perceive the robustness of the vehicle. In the real test of outer panel stiffness after prototype production, evaluators manually press the outer panel in advance to identify vulnerable areas to be tested and evaluate the performance only in those area. However, when developing the outer panel stiffness performance using FEA (Finite Element Analysis) before releasing the drawing, it is not possible to filter out these areas, so the entire outer panel must be evaluated. This requires a significant amount of computing resources and manpower. In this study, an approach utilizing artificial intelligence was proposed to streamline the outer panel stiffness analysis and improve development reliability.

After the COVID-19 pandemic, leisure activities and cultures have undergone significant transformations. Particularly, there has been an increased demand for outdoor camping. Consequently, the need for capabilities that allow vehicles to navigate not only paved roads but also unpaved and rugged terrains has arisen. In this study, we aim to address this demand by utilizing AI to introduce a 'Stuck Probability Estimation Algorithm' for vehicles on off-road. To estimate the 'Stuck Probability' of a vehicle, a mathematical model representing vehicle behavior is essential. The behavior of off-road driving vehicles can be characterized in two main aspects: firstly, the harshness of the terrain (how uneven and rugged it is), and secondly, the extent of wheel slip affecting the vehicle's traction.

Testing of ducted fuel injection (DFI) in a single-cylinder engine with production-like hardware previously showed that adding a duct structure increased soot emissions at the full load, rated speed operating point [1]. The authors hypothesized that the DFI flame, which travels faster than a conventional diesel combustion (CDC) flame, and has a shorter distance to travel, was being re-entrained into the on-going fuel injection around the lift-off length (LOL), thus reducing air entrainment into the on-going injection. The engine operating condition and the engine combustion chamber geometry were duplicated in a constant pressure vessel. The experimental setup used a 3D piston section combined with a glass fire deck allowing for a comparison between a CDC flame and a DFI flame via high-speed imaging. CH* imaging of the 3D piston profile view clearly confirmed the re-entrainment hypothesis presented in the previous engine work.

We are in the context of the analysis of carbon fiber reinforced plastics (CFRP) high-pressure vessel (COPV - Composite Overwrapped Pressure Vessel) manufactured by filament winding (FW). Classically, the parameters of material models are identified based on flat laminate coupons with specific predetermined fiber orientations, and based on standards like the ones of ASTM relevant for flat coupons. CFRP manufactured by FW has a unique and complex laminate structure, which presents curvatures and ply interlacements. In practice, it is important to use coupons produced with the final manufacturing process for the parameter identification of the material models; if classical coupons produced by e.g. ply lamination are used, the effect of FW structure cannot be accounted for, and cannot be introduced in the material models. It is therefore essential to develop an approach to create representative flat coupons based on the FW process.

The tripod constant velocity joint (CVJ) has been widely used in mechanical systems due to its strong load-bearing capacity, high efficiency, and reliability. It has become the most commonly used plunging-type CVJ in automotive drive-shaft. A generated axial force (GAF) with a third-order characteristic of driven shaft speed is caused by the internal friction and motion characteristics in a tripod joint. The large GAF has a negative impact on the NVH (Noise, Vibration, and Harshness) characteristics of automobiles, and this issue is particularly prominent in new energy vehicles. A multi-body dynamic model of the Adjustable Angular Roller (AAR) tripod CVJ is developed to calculate and analyze the GAF. To describe the internal motion of the AAR tripod CVJ, the contact interactions between the roller and the track or the trunnion were modeled using non-linear equivalent spring-damping models for contact collision forces and modified Coulomb friction model for friction.

Heavy Vehicle Event Data Recorders (HVEDRs) have the ability to capture important data surrounding an event such as a crash or near crash. Efforts by many researchers to analyze the capabilities and performance of these complex systems can be problematic, in part, due to the challenges of obtaining a heavy truck, the necessary space to safely test systems, the inherent unpredictability in testing, and the costs associated with this research. In this paper, a method for simulating vehicle speed sensor (VSS) inputs to HVEDRs to trigger events is introduced and validated. Full-scale instrumented testing is conducted to capture raw VSS signals during steady state and braking conditions. The recorded steady state VSS signals are injected into the HVEDR along with synthesized signals to evaluate the response of the HVEDR. Brake testing VSS signals are similarly captured and injected into the HVEDR to trigger an event record.

The Advanced Driver Assistance System (ADAS) is a comprehensive feature set designed to aid a driver in avoiding or reducing the severity of collisions while operating the vehicle within specified conditions. In General Motors (GM) vehicles, the primary controller for the ADAS is the Active Safety Control Module (ASCM). In the 2013 model year, GM introduced an ASCM utilizing the GM internal nomenclature of External Object Calculation Module (EOCM) in some of their vehicles produced for the North American market. Similar to the Sensing and Diagnostic Module (SDM) utilized in the restraints system, the EOCM3 LC contains an Event Data Recorder (EDR) function to capture and record information surrounding certain ADAS or Supplemental Inflatable Restraint (SIR) events. The ASCM EDR contains information from external object sensors, various chassis and powertrain control modules, and internally calculated data.

When investigating traffic accidents, it is important to determine the causes. To do so, it is necessary to reconstruct the accident situation accurately and in detail using objective and diverse information. We propose a method for reconstructing the accident situation (“reconstruction method”) which consists of rebuilding the situation immediately before the collision (“pre-crash situation”) using data collected during that time by an event data recorder (EDR) and a dashboard camera (DBC) onboard one or both of the vehicles involved. First, the vehicle’s traveling trajectory was integrally calculated using the vehicle speed and yaw rate recorded by the EDR, each point along the trajectory being linked to the EDR data.

Vehicle efficiency and range, along with the DC charging speed, are deemed as the most important criteria for an electric vehicle currently. The electric vehicle energy consumption is impacted by the change in temperature along with the driving style and average speed of a customer, all other factors being constant. Hence understanding the patterns and impact of different aspects of an EV range & charging speed is crucial in delivering an electric vehicle with robust efficiency across all weather conditions. In this paper we have analysed vehicle parameters of global Jaguar I-PACE customer data. We present and analyse the collated big data of around 50,000+ unique vehicles with a data aggregate of well over 482 million km. In moderate ambient conditions the analysis indicated a good correlation with 50th to 75th percentile drivers’ energy consumption to the EPA label figure.

Android applications have historically faced vulnerabilities to man-in-the-middle attacks due to insecure custom SSL/TLS certificate validation implementations. In response, Google introduced the Network Security Configuration (NSC) as a configuration-based solution to improve the security of certificate validation practices. NSC was initially developed to enhance the security of Android applications by providing developers with a framework to customize network security settings. However, recent studies have shown that it is often not being leveraged appropriately to enhance security. Motivated by the surge in vehicular connectivity and the corresponding impact on user security and data privacy, our research pivots to the domain of mobile applications for vehicles. As vehicles increasingly become repositories of personal data and integral nodes in the Internet of Things (IoT) ecosystem, ensuring their security moves beyond traditional issues to one of public safety and trust.

The operation management of electric Taxi fleets requires cooperative optimization of Charging and Dispatching. The challenge is to make real-time decisions about which is the optimal charging station or passenger for each vehicle in the fleet. With the rapid advancement of Vehicle Internet of Things (VIOT) technologies, the aforementioned challenge can be readily addressed by leveraging big data analytics and machine learning algorithms, thereby contributing to smarter transportation systems. This study focuses on optimizing real-time decision-making for charging and dispatching in large-scale electric taxi fleets to improve their long-term benefits. To achieve this goal, a spatiotemporal decision framework using Bi-level optimization is proposed. Initially, a deep reinforcement learning-based model is built to estimate the value of charging and order dispatching under uncertainty.