Search Results

Accurately predicting the future trajectories of surrounding traffic agents is important for ensuring the safety of autonomous vehicles. To address the scenario of frequent interactions among traffic agents in the highway merging area, this paper proposes a trajectory prediction method based on interactive graph attention mechanism. Our approach integrates an interactive graph model to capture the complex interactions among traffic agents as well as the interactions between these agents and the contextual map of the highway merging area. By leveraging this interactive graph model, we establish an agent-agent interactive graph and an agent-map interactive graph. Moreover, we employ Graph Attention Network (GAT) to extract spatial interactions among trajectories, enhancing our predictions. To capture temporal dependencies within trajectories, we employ a Transformer-based multi-head self-attention mechanism.

The tooth surface error will affect the contact pattern and transmission error of the hypoid gear, which may result in an unfavorable dynamic response. The tooth surface error can be generated by machine tool errors, such as blade wear. The most common forms of blade wear are the positive cutter radius and the positive blade angle error. In addition, in the cutting process of face-hobbed hypoid gear, the continuous indexing motion will aggravate the blade wear due to the alternating cutting force. Most previous studies on the influence of hypoid gear tool errors only focus on the contact pattern and static transmission error. However, there are very few studies about the effect of tool errors on hypoid gear dynamic responses. In this paper, a hypoid gear tooth surface, mesh, and linear dynamic model with tool errors were established. The tooth surface deviation distribution of different tool errors was analyzed.

Deep reinforcement learning (DRL) has not been widely used in the engineering field yet because RL needs to be learned through ‘trial and error’, which makes the application of this kind of algorithm in real physical environment more difficult, and it is impossible to carry out ‘trial and error’ learning on real vehicles. By analyzing the motion state of the vehicle in the car following mode, the algorithm that combined traditional longitudinal motion control with DRL improves the safety of RL in the real physical environment and the poor adaptability of the traditional longitudinal motion control algorithm. In this paper, the longitudinal motion of the unmanned vehicle is taken as the research object, and the PID algorithm is combined with the Deep Deterministic Policy Gradient (DDPG) algorithm to control the longitudinal motion of the unmanned vehicle.

Motor thermal management of electric vehicles (EVs) is becoming more significant due to its close relations to vehicle aerodynamic performance and power consumption, while computer aided engineering (CAE) plays an important role in its development. A 1D-3D coupled model is established to characterize transient thermal performance of the motor in an electric vehicle on a high performance computer (HPC) platform. The 1D motor thermal management model is integrated with the 1D powertrain model, and a 3D thermal model is established for the motor, while online data exchange is realized between the 1D and 3D models. The 1D model gives boundaries such as inlet coolant temperature, mass flowrate and motor heat generation to the 3D model, while the 3D model gives back boundaries such as heat transfer to coolant simultaneously. Transient simulations are performed for the 140kph(20°C) driving cycle, and the model is calibrated with experimental data.

In order to improve the performance of automatic emergency steering and collision avoidance of intelligent vehicle, two automatic steering control methods under ideal model following control are proposed. The two ideal reference models are the reference model with zero sideslip angle of vehicle gravity center and the reference model with no phase-lag in vehicle lateral acceleration. The control system adopts the combination of outer loop and inner loop. In the design of the outer loop controller, the optimal control is used to get the steering wheel angle needed to avoid collision. The inner loop controller uses feedforward and feedback control to get the required front and rear wheel steering angles. Taking vehicle two degrees of freedom (DOF) lateral dynamics model as the research object, the vehicle collision avoidance reference trajectory is obtained through the fifth-degree polynomial.

The objective of this study was to assess the formability of two 3rd generation advanced high strength steels (3rd Gen AHSS) with ultimate strengths of 980 and 1180 MPa and evaluate their applicability to a structural B-Pillar for a mid-sized sport utility vehicle. The constitutive behavior including strain-rate effects and formability were characterized to generate the material models for use within AutoForm R8 software to design the B-pillar tooling and forming process. An extended Bressan-Williams instability model was able to deterministically predict the forming limit curves obtained using Marciniak tests. The tooling for the representative B-pillar was designed and fabricated with Bowman Precision Tooling and forming trials conducted for both 3rd Gen steels that had a thickness of 1.4 mm.

Forecasting the motion of the leading vehicle is a critical task for connected autonomous vehicles as it provides an efficient way to model the leading-following vehicle behavior and analyze the interactions. In this study, a personalized time-series modeling approach for leading vehicle trajectory prediction considering different driving styles is proposed. The method enables a precise, personalized trajectory prediction for leading vehicles with limited inter-vehicle communication signals, such as vehicle speed, acceleration, space headway, and time headway of the front vehicles. Based on the learning nature of human beings that a human always tries to solve problems based on grouping and similar experience, three different driving styles are first recognized based on an unsupervised clustering with a Gaussian Mixture Model (GMM).

Residual stress prediction arising from manufacturing processes provides paramount information for the fatigue performance assessment of components subjected to cyclic loading. The determination of the material model to be applied in the numerical model should be taken carefully. This study focuses on the estimation of residual stresses generated after deep rolling of cast iron crankshafts. The researched literature on the field employs the available commercial material codes without closer consideration on their reverse loading capacities. To mitigate this gap, a single element model was used to compare potential material models with tensile-compression experiments. The best fit model was then applied to a previously developed crankshaft deep rolling numerical model. In order to confront the simulation outcomes, residual stresses were measured in two directions on real crankshaft specimens that passed through the same modeled deep rolling process.

With the widespread development of automated driving systems (ADS), it is imperative that standardized testing methodologies be developed to assure safety and functionality. Scenario testing evaluates the behavior of an ADS-equipped subject vehicle (SV) in predefined driving scenarios. This paper compares four modes of performing such tests: closed-course testing with real actors, closed-course testing with surrogate actors, simulation testing, and closed-course testing with mixed reality. In a collaboration between the Waterloo Intelligent Systems Engineering (WISE) Lab and AAA, six automated driving scenario tests were executed on a closed course, in simulation, and in mixed reality. These tests involved the University of Waterloo’s automated vehicle, dubbed the “UW Moose”, as the SV, as well as pedestrians, other vehicles, and road debris.

1 Crack initiation and propagation test data gathered during tests on Keyhole notched samples is used to evaluate a fatigue life prediction technique. Materials tested include a lower strength ManTen steel and a higher strength Boron steel, RQC-100, both tested with constant and variable amplitude histories. Initiation fatigue life is predicted using the usual method of plasticity correction at the notch followed by a Palmgren-Miner summation of damage with mean stress correction. The emphasis of the study is on simulating the crack propagation results. For that phase discretetize da/dN vs ΔK lines and thresholds for negative R ratios, are used specifically to help predict the propagation for one of the VA histories that had a significant negative mean. The open source crack propagation simulation program applies a material memory model to determine the crack advance on a reversal by reversal basis.

The vehicle axle gear whine noise and vibration are key issues for the automotive industry to design a quiet, reliable driveline system. The main source of excitation for this vibration energy comes from hypoid gear transmission error (TE). The vibration transmits through the flexible axle components, then radiates off from the surface of the housing structure. Thus, the design of hypoid gear pair with minimization of TE is one way to control the dynamic behavior of the vehicle axle system. In this paper, an approach to obtain minimum TE and improved dynamic response with optimal tooth profile modification parameters is discussed. A neural network algorithm, named Back Propagation (BP) algorithm, with improved Particle Swarm Optimization (PSO) is used to predict the TE if some tooth profile modification parameters are given to train the model.

In this paper, a new solution method is presented to study the effect of wave propagation in engine manifolds, which includes solving one-dimensional models for compressible flow of air. Velocity, pressure, and density profiles are found by solving a system of non-linear Partial Differential Equations (PDEs) in space and time derived from Euler’s equations. The 1D model includes frictional losses, area change, and heat transfer. The solution is traditionally found by utilizing the Method of Characteristics and applying finite difference solutions to the resulting system of ordinary differential equations (ODEs) over a discretized grid. In this work, orthogonal collocation is used to solve the system of ODEs that is defined along the characteristic curves. Orthogonal polynomials are utilized to approximate velocity, pressure, sound speed, and the characteristic curves along which the system of PDEs reduce to a system of ODEs.

Fatigue crack initiation and propagation models predict the fatigue life of welded "T" specimens tested by the Fatigue Design and Evaluation (FDE) Committee of SAE under constant and variable amplitude load histories. The crack propagation equations stipulated by British Standard BS-7910 have been incorporated in a material memory model for cyclic deformation. The simulations begin with the crack initiation model and show how it is used to account for cyclic mean stress relaxation and the effects of periodic overloads. After the cracks initiate the BS-7910 model is applied to predict the crack advance due to either constant or variable amplitude histories. Simulation results correspond to the experimental results with good accuracy.

Most of algorithms of lane detection mainly aim at the scenes of daytime. However, those algorithms are unstable for the lane detection at night because the camera is very sensitive to the light change. This paper proposed a lane detection algorithm that largely improves the detection system’s performance when it is used at night. The algorithm has two main stage: Image processing and Kalman filter (KF). The key process steps of Stage 1 are: extracting the Region of Interesting (ROI)→Edge Detection →Binarization→Hough→ Lane Selection→Lane fitting. First step, a ROI could be extracted according to the relatively fixed location of lanes. In step of edge detection, we use a creative filter named Correlation filter to remove image noise and remain the feature of lane. The filter matrix looks like “[0 1 1, −1 0 1; −1 −1 0]”. Next, the candidate lines are detected by the Hough transform, then, the equations of lane are acquired by fitting spots obtained from Hough.

One-dimensional (1D) simulation tools, the computing speed of which is relatively fast, usually solve simple complexity problems. The solving process of 1D simulation is mostly based on one-dimensional dynamic equations and empirical laws and thus in some cases it cannot obtain a similar accuracy with the time-consuming three-dimensional (3D) simulation tools. The 1D-3D co-simulation, which combines the advantages of the two simulation tools while minimizes the disadvantages, is a method that integrates and runs the two simulation tools concurrently. The coupled simulation can offer a 3D analysis for which a detailed information is needed while offer system level information in the rest of the whole system where averaged results are sufficient. The approach not only minimizes the computational cost, but avoids demand for imposing accurate boundary conditions to the 3D simulation.

This paper discusses modeling of a power-split hybrid electric vehicle and the design of a longitudinal dynamics controller for the University of Waterloo’s self-driving vehicle project. The powertrain of Waterloo’s vehicle platform, a Lincoln MKZ Hybrid, is controlled only by accelerator pedal actuation. The vehicle’s power management strategy cannot be altered, so a novel approach to grey-box modeling of the OEM powertrain control architecture and dynamics was developed. The model uses a system of multiple neural networks to mimic the response of the vehicle’s torque control module and estimate the distribution of torque between the powertrain’s internal combustion engine and electric motors. The vehicle’s power-split drivetrain and longitudinal dynamics were modeled in MapleSim, a modeling and simulation software, using a physics-based analytical approach.

In this paper, gear ratios of a two-speed transmission system are optimized for an electric passenger car. Quasi static system models, including the vehicle model, the motor, the battery, the transmission system, and drive cycles are established in MATLAB/Simulink at first. Specifically, since the regenerative braking capability of the motor is affected by the SoC of battery and motors torque limitation in real time, the dynamical variation of the regenerative brake efficiency is considered in this study. To obtain the optimal gear ratios, iterations are carried out through Nelder-Mead algorithm under constraints in MATLAB/Simulink. During the optimization process, the motor efficiency is observed along with the drive cycle, and the gear shift strategy is determined based on the vehicle velocity and acceleration demand. Simulation results show that the electric motor works in a relative high efficiency range during the whole drive cycle.

The development of a comprehensive battery simulator is essential for future improvements in the durability, performance and service life of lithium-ion batteries. Although simulations can never replace actual experimental data, they can still be used to provide valuable insights into the performance of the battery, especially under different operating conditions. In addition, a single-cell model can be easily extended to the pack level and can be used in the optimization of a battery pack. The first step in building a simulator is to create a model that can effectively capture both the voltage response and thermal behavior of the battery. Since these effects are coupled together, creating a robust simulator requires modeling both components. This paper will develop a battery simulator, where the entire battery model will be composed of four smaller submodels: a heat generation model, a thermal model, a battery parameter model and a voltage response model.

In this paper, a degradation testing of a lithium-ion battery used for an electric vehicle (EV) is performed and the capacity fade is measured over 400 cycles. For this, a 200 Ah LiFePO4 battery cell is tested under ambient temperature conditions with charge-discharge cycles at rate of 1C (constant current). Additionally, individual cell characterization is conducted using a C/25 (0.8A) charge-discharge cycle and hybrid pulse power characterization (HPPC). Later, the Thevenin battery model was constructed in MATLAB along with an empirical degradation model and validated in terms of voltage for all cycles. It is also found that the presented model closely estimated the profiles observed in the experimental data. Data collected from the experimental results showed that a capacity fade occurred over the 400 cycles and the discharge capacity at the end of 400th cycle is found to be 137.73 Ah. The error between model/experiments is found to be less than 3.5% for all cycles.

The interactions between automatic controls, physics, and driver is an important step towards highly automated driving. This study investigates the dynamical interactions between human-selected driving modes, vehicle controller and physical plant parameters, to determine how to optimally adapt powertrain control to different human-like driving requirements. A cyber-physical system (CPS) based framework is proposed for co-design optimization of the physical plant parameters and controller variables for an electric powertrain, in view of vehicle’s dynamic performance, ride comfort, and energy efficiency under different driving modes. System structure, performance requirements and constraints, optimization goals and methodology are investigated. Intelligent powertrain control algorithms are synthesized for three driving modes, namely sport, eco, and normal modes, with appropriate protocol selections. The performance exploration methodology is presented.