Vehicle Dynamics Research Articles

A hybrid physics-data driven approach for vehicle dynamics state estimation

A hybrid physics-data driven approach for vehicle dynamics state estimation

A comparison of reinforcement learning policies for dynamic vehicle routing problems with stochastic customer requests

This paper presents directions for using reinforcement learning with neural networks for dynamic vehicle routing problems (DVRPs). DVRPs involve sequential decision-making under uncertainty where the expected future consequences are ideally included in current decision-making. A frequently used framework for these problems is approximate dynamic programming (ADP) or reinforcement learning (RL), often in conjunction with a parametric value function approximation (VFA). A straightforward way to use VFA in DVRP is linear regression (LVFA), but more complex, non-linear predictors, e.g., neural network VFAs (NNVFA), are also widely used. Alternatively, we may represent the policy directly, using a linear policy function approximation (LPFA) or neural network PFA (NNPFA). The abundance of policies and design choices complicate the use of neural networks for DVRPs in research and practice. We provide a structured overview of the similarities and differences between the policy classes. Furthermore, we present an empirical comparison of LVFA, LPFA, NNVFA, and NNPFA policies. The comparison is conducted on several problem variants of the DVRP with stochastic customer requests. To validate our findings, we study realistic extensions of the stylized problem on (i) a same-day parcel pickup and delivery case in the city of Amsterdam, the Netherlands, and (ii) the routing of robots in an automated storage and retrieval system (AS/RS). Based on our empirical evaluation, we provide insights into the advantages and disadvantages of neural network policies compared to linear policies, and value-based approaches compared to policy-based approaches.

Regression algorithm based residual range prediction and validation on EV travel data

ABSTRACT Range anxiety is one of the daunting facts that holds back customers from purchasing electric vehicles (EVs). The range of an EV depends on many internal and external parameters that cause large deviations in the energy left in the battery. taxi and bus fleets also decide on a fixed distance of travel after which they recharge the battery so as to avoid unwanted last-minute surprises. In a country like India, range prediction methods have to be extensively worked out due to inadequate charging infrastructure and vivid range of climatic and terrain conditions. This paper presents residual range prediction methods implemented in two different models of EVs based on real travel data on Indian roads. The data collected are from two compact Sports Utility Vehicles (SUV) in the Indian automobile sector, namely MGZSEV and TATA Nexon EV. The travel data of MGZSEV was collected from the mobile application whereas the data collected from Tata Nexon EV was from the dashboard display of the vehicle. Two different approaches were implemented, first being the conventional vehicle dynamics based mathematical method and second using Machine Learning (ML) based algorithms. The analysis and plots are discussed in detail and interpreted.

Large Language Models (LLMs) as Traffic Control Systems at Urban Intersections: A New Paradigm

This study introduces a novel approach for traffic control systems by using Large Language Models (LLMs) as traffic controllers. The study utilizes their logical reasoning, scene understanding, and decision-making capabilities to optimize throughput and provide feedback based on traffic conditions in real time. LLMs centralize traditionally disconnected traffic control processes and can integrate traffic data from diverse sources to provide context-aware decisions. LLMs can also deliver tailored outputs using various means such as wireless signals and visuals to drivers, infrastructures, and autonomous vehicles. To evaluate LLMs’ ability as traffic controllers, this study proposed a four-stage methodology. The methodology includes data creation and environment initialization, prompt engineering, conflict identification, and fine-tuning. We simulated multi-lane four-leg intersection scenarios and generated detailed datasets to enable conflict detection using LLMs and Python simulation as a ground truth. We used chain-of-thought prompts to lead LLMs in understanding the context, detecting conflicts, resolving them using traffic rules, and delivering context-sensitive traffic management solutions. We evaluated the performance of GPT-4o-mini, Gemini, and Llama as traffic controllers. Results showed that the fine-tuned GPT-mini achieved 83% accuracy and an F1-score of 0.84. The GPT-4o-mini model exhibited a promising performance in generating actionable traffic management insights, with high ROUGE-L scores across conflict identification of 0.95, decision making of 0.91, priority assignment of 0.94, and waiting time optimization of 0.92. This methodology confirmed LLMs’ benefits as a traffic controller in real-world applications. We demonstrated that LLMs can offer precise recommendations to drivers in real time including yielding, slowing, or stopping based on vehicle dynamics. This study demonstrates LLMs’ transformative potential for traffic control, enhancing efficiency and safety at intersections.

Research on a Hierarchical Control Strategy for Anti-Lock Braking Systems Based on Active Disturbance Rejection Control (ADRC)

To improve the slip rate control effect for different road conditions during emergency braking of wheel hub motor vehicles, as well as to address the problems of uncertainty and nonlinearity of the system when the electro-mechanical braking system is used as the actuator of the ABS, a hierarchical control strategy of the anti-lock braking system (ABS) using active disturbance rejection control (ADRC) is proposed. Firstly, a vehicle dynamics model and an ABS model based on the EMB system are established; secondly, a speed observer based on the dilated state observer is used in the upper layer to design a pavement recognition algorithm, which recognizes the current pavement and outputs the optimal slip rate; then, an ABS controller based on the ADRC algorithm is designed for the lower layer to track the optimal slip rate. In order to verify the performance of the pavement recognition method and control strategy, vehicle simulation software is used to establish the model and simulation. The results show that the road surface recognition method can quickly and effectively recognize the road surface, and comparing the emergency braking control effects of PID and SMC under different road surface conditions, the ADRC strategy has better robustness and reliability, and improves the braking effect.

Quantification Method of Driving Risks for Networked Autonomous Vehicles Based on Molecular Potential Fields

Connected autonomous vehicles (CAVs) face constraints from multiple traffic elements, such as the vehicle, road, and environmental factors. Accurately quantifying the vehicle’s operational status and driving risk level in complex traffic scenarios is crucial for enhancing the efficiency and safety of connected autonomous driving. To continuously and dynamically quantify the driving risks faced by CAVs in the road environment—arising from the front, rear, and lateral directions—this study focused s on the self-driving particle characteristics that enable CAVs to perceive their surrounding environment and make driving decisions. The vehicle-to-vehicle interaction behavior was analogized to the inter-molecular interaction relationship, and a molecular Morse potential model was applied, coupled with the vehicle dynamics theory. This approach considers the safety margin and the specificity of driving styles. A multi-layer decoder–encoder long short-term memory (LSTM) network was employed to predict vehicle trajectories and establish a risk quantification model for vehicle-to-vehicle interaction behavior. Using SUMO software (win64-1.11.0), three typical driving behavior scenarios—car-following, lane-changing, and yielding—were modeled. A comparative analysis was conducted between the risk field quantification method and existing risk quantification indicators such as post-encroachment time (PET), deceleration rate to avoid crash (DRAC), modified time to collision (MTTC), and safety potential fields (SPFs). The evaluation results demonstrate that the risk field quantification method has the advantage of continuously quantifying risk, addressing the limitations of traditional risk indicators, which may yield discontinuous results when conflict points disappear. Furthermore, when the half-life parameter is reasonably set, the method exhibits more stable evaluation performance. This research provides a theoretical basis for the dynamic equilibrium control of driving risks in connected autonomous vehicle fleets within mixed-traffic environments, offering insights and references for collision avoidance design.

Correspondence Between Alignment Wavelength and Vehicle Response Wavelength of High-Speed Rails

The acceleration of the vehicle body is closely related to the wavelength components of the track, making the design of track alignments crucial for enhancing the comfort of high-speed rail travel. However, research on the effects of long-wavelength deformations on alignment design and their impact on vehicle dynamics remains limited, and existing methodologies fail to effectively calculate the wavelengths of traditional alignments, which consist of linear segments and vertical curves. This study introduces a novel approach for computing long-wavelength track alignments and vehicle response wavelengths using signal decomposition and the Hilbert transform. A comprehensive program is developed to analyze the relationship between alignment and vehicle response wavelengths through correlation coefficients and coherence functions. The results show that traditional alignments have predominant wavelengths exceeding 1 000[Formula: see text]m, which exhibit negligible correlation with vehicle response due to the dominant influence of alignment curvature. In contrast, Fourier series-designed alignments demonstrate a significant correlation with vehicle response, particularly under high-speed conditions. Quantitatively, vehicle response wavelengths are found to range from 1[Formula: see text]000[Formula: see text]m to 5[Formula: see text]000[Formula: see text]m, with Fourier series alignments effectively reducing dynamic impacts. These findings highlight the potential of the proposed approach in optimizing track design for challenging railway sections, providing valuable guidance for improving high-speed railway operations.

Stability analysis of linear single-track models with transient tyre dynamics

The present paper investigates the behaviour of linear single-track vehicle models considering transient tyre dynamics. In particular, the tyre models covered in this paper include lumped-parameter approximations, like the famous single-contact point and two-regime formulations, as well as refined distributed brush-like representations formulated in terms of partial differential equations (PDEs). In the first case, the resulting description consists of a system of linear ordinary differential equations (ODEs), whereas, in the second case, the lateral vehicle dynamics are captured by an interconnection of ODEs and PDEs. Stability analyses are conducted concerning both variants. In particular, for single-track models equipped with lumped tyre approximations, sufficient conditions on the vehicle's parameters that ensure dynamical stability are explicitly derived. Concerning the more involved formulation that considers the exact distributed brush model, the exact criterion for stability is instead deduced analytically, and then a qualitative analysis is performed. Via a number of examples, it is revealed that, for a wide range of frequencies of the steering input, the vehicle's behaviour predicted by the two models is similar to that obtained using the classic single-track formulation. The findings contained in the paper may assist vehicle manufacturers in setting requirements on design parameters, and in developing control algorithms that account for transient tyre dynamics at different degrees of accuracy.

Road preview method for active suspension based on reinforcement learning

Abstract The pre-identification of impulse road surfaces, such as speed bumps, potholes, and manhole covers, can significantly enhance the performance of active suspension systems. However, existing current identification methods often fail to balance between low cost, high reliability, and precise speed. This study proposes an impulse road surface identification method based on a reinforcement learning network. First, a real dataset of impulse road surfaces was collected, with suspension dynamic responses serving as inputs and random road levels along with impulse road surface features as outputs. This data was used to develop a Random Forest Extreme Gradient Boosting (RF-XGBoost) network to recognize road surface information from vehicle dynamics. Next, a semantic segmentation network was employed to segment road surfaces during intelligent vehicle travel, using road surface information identified by the random forest network as the reward function for reinforcement learning. The reinforcement learning policy was pre-trained using the actual collected data. To validate the effectiveness of the proposed control strategy, a simulation environment was constructed in Prescan, where speed bumps, potholes, and manhole covers of varying heights and sizes were randomly arranged. The reinforcement learning-based road surface identification algorithm was implemented in Simulink, and a co-simulation was conducted with the CarSim vehicle model. The enhanced RF-XGBoost network effectively distinguishes between random and impulse road surfaces, achieving average recognition accuracies of 95.7% for random surfaces and 98.2% for impulse surfaces. During initial training iterations, the RL network exhibited lower accuracy; however, as training progressed, the accuracy for unfamiliar road surface features reached an average of 96.5%, and overall recognition accuracy improved by an average of 9%. The simulation results demonstrate that the proposed reinforcement learning identification method effectively acquires information about the road ahead, shows robustness and generalization, and provides crucial disturbance data for subsequent active suspension control.

Real-time Adversarial Image Perturbations for Autonomous Vehicles using Reinforcement Learning

The deep neural network (DNN) model for computer vision tasks (object detection and classification) is widely used in autonomous vehicles, such as driverless cars and unmanned aerial vehicles. However, DNN models are shown to be vulnerable to adversarial image perturbations. The generation of adversarial examples against inferences of deep neural networks has been actively studied recently. The generation typically relies on optimizations taking an entire image frame as the decision variable. Hence, given a new image, the computationally expensive optimization needs to start over as there is no learning between the independent optimizations. Very few approaches have been developed for attacking online image streams while taking into account the underlying physical dynamics of autonomous vehicles, their mission, and the environment. The paper presents a multi-level reinforcement learning framework that can effectively generate adversarial perturbations to misguide autonomous vehicles’ missions. In the existing image attack methods against autonomous vehicles, optimization steps are repeated for every image frame. This framework removes the need for fully converged optimization at every frame. Using multi-level reinforcement learning, we integrate a state estimator and a generative adversarial network that generates the adversarial perturbations. Due to the reinforcement learning agent consisting of state estimator, actor, and critic that only uses image streams, the proposed framework can misguide the vehicle to increase the adversary’s reward without knowing the states of the vehicle and the environment. Simulation studies and a robot demonstration are provided to validate the proposed framework’s performance.

Analytical solution of vehicle phase-plane equilibria through the Root-Rational tyre model

Passenger vehicle steady-state behaviour has been investigated for a long time. Among the existing methods, the phase portrait is rather popular and useful. However, the computation of vehicle equilibria may only be done numerically. Furthermore, the numerical procedure must be repeated in case of changes in the inputs (e.g. steering angle). This paper proposes a new methodology to obtain an analytical solution of vehicle equilibria. The key is the definition of a new tyre model, denoted as Root-Rational (RR) tyre model, which allows inverting the vehicle dynamics equations in closed form. Using the proposed tyre model reduces the equilibria location problem to finding the roots of a third-order polynomial. After describing the procedure in detail, comparisons are made between the analytical solution and a classical numerical approach, either using the same RR tyre model or a more classical one. Results show great accuracy when locating the equilibria coordinates and a significantly reduced computational time, paving the way for innovative vehicle stability control algorithms based on the real-time identification of the stability region in the phase plane.

Online Identification and Detection of Manual Control Adaptation with Recursive Time Series

An online pilot manual control behavior identification method, based on recursive low-order time-series model estimation, is presented and validated using experimental data. Eight participants performed compensatory tracking tasks with time-varying vehicle dynamics, where, at an unpredictable moment during a run, a sudden degradation in dynamics could occur. They were instructed to push a button when they detected a change in dynamics. Two methods to automatically detect the moment when pilot adaptation occurs from online estimated parameter traces are discussed. Results show that pilots are more accurate in detecting changes than either algorithm. But when the algorithms are correct, they are often quicker to detect pilot adaptation than pilots themselves. The presented techniques have potential but need improvements.

Height Control Method for Air Suspension Systems Based on Model-Free Adaptive Control and Improved Genetic Algorithm

<div>This article presents a height control method for air suspension systems, which are influenced by strong nonlinearity and multiple coupling factors, based on model-free adaptive control (MFAC) using full-form dynamic linearization (FFDL). To address the impact of different damping coefficients of the shock absorber on the height control effect, an improved genetic algorithm is employed to globally optimize the relevant parameters involved in the design of the control law, thereby enhancing the height control performance. The precision of modeling the air suspension system has a direct impact on the simulation of both static and dynamic vehicle models, as well as the accuracy of height control. In this article, an equivalent thermodynamic model of the air suspension system is established based on the principle of energy conservation for height control research. Considering the nonlinearity of the air suspension system and the need to make additional assumptions before modeling, a MFAC method using FFDL is adopted for controller design. Traditional height control methods do not consider the impact of changes in the shock absorber damping coefficient on the height control effect. For different damping coefficients, the body height tracking error is large when using the same height control law initialization parameters. Therefore, an improved genetic algorithm is employed to globally optimize the MFAC parameters under different damping states. The effectiveness of the thermodynamic model of the air suspension system and the MFAC method for height control, with parameters tuned using the improved genetic algorithm, was validated through MATLAB/Simulink simulations.</div>

"Simulation Approach to Investigate the Influence of the Steering Wheel Angle and Vehicle Speed on the Rollover of a 4-Axle Truck Vehicle during Turning Maneuvers"

The article aims to investigate the influence of vehicle speed and steering wheel angle on the rollover condition of a 4-axle truck vehicle. A 30-degree-of-freedom (DOF) dynamic model of a 4-axle truck is established using the Multibody System Method and the Newton-Euler equation system. The vehicle body is described with 6 DOFs. Additionally, 2 DOFs are used to describe the roll motion of the front and rear masses of the vehicle body, taking into account the influence of the torsional stiffness of the vehicle frame on vehicle dynamics. The Burckhardt tire model calculates the interaction forces between tires and the road, using experimental coefficients corresponding to the road with a maximum adhesion coefficient of 0.8. The simulation uses MATLAB-Simulink, with a 5 to 65 km/h speed range and a steering wheel rotation angle ranging from 25 to 300 degrees. The results have determined the vehicle rollover conditions based on the signs of the load transfer ratio of the axle and the entire vehicle. Together with the results of lateral acceleration, yaw rate, and roll angle of the vehicle. This result can be used to propose dynamic thresholds for designing early warning systems and anti-rollover control systems for multi-axle truck vehicles.

A State Estimation of Dynamic Parameters of Electric Drive Articulated Vehicles Based on the Forgetting Factor of Unscented Kalman Filter with Singular Value Decomposition

In this paper, a state estimation method of distributed electric drive articulated vehicle dynamics parameters based on the forgetting factor unscented Kalman filter with singular value decomposition (SVD-UKF) is proposed. The 7DOF nonlinear dynamics model of a distributed electric drive articulated vehicle is established. The unscented Kalman filter algorithm is the foundation, with singular value decomposition replacing the Cholesky decomposition. A forgetting factor is introduced to dynamically adapt the observation noise covariance matrix in real time, resulting in a centralized parameter state estimator for the articulated vehicle. The proposed parameter state estimation method based on the forgetting factor SVD-UKF is simulated and compared with the unscented Kalman filter (UKF) estimation method. Key dynamic parameters are estimated, such as the lateral and longitudinal velocities and accelerations, angular velocity, articulated angle, wheel speeds, and longitudinal and lateral tire forces of both the front and rear vehicle bodies. The results show that the proposed forgetting factor SVD-UKF method outperforms the traditional UKF method. Furthermore, a prototype vehicle test is conducted to compare the estimated values of various dynamic parameters with the actual values, demonstrating minimal errors. This verifies the effectiveness of the proposed dynamic parameter estimation method for articulated vehicles.

A neighborhood search-based heuristic for the dynamic vehicle routing problem

A neighborhood search-based heuristic for the dynamic vehicle routing problem

Stability Study of Distributed Drive Vehicles Based on Estimation of Road Adhesion Coefficient and Multi-Parameter Control

In order to improve the driving stability of distributed-drive intelligent electric vehicles under different roadway attachment conditions, this paper proposes a multi-parameter control algorithm based on the estimation of road adhesion coefficients. First, a seven-degree-of-freedom (7-DOF) vehicle dynamics model is established and optimized with a layered control strategy. The upper-level control module calculates the desired yaw rate and sideslip angle using the two-degree-of-freedom (2-DOF) vehicle model and estimates the road adhesion coefficient by using the singular-value optimized cubature Kalman filtering (CKF) algorithm; the middle-level utilizes the second-order sliding mode controller (SOSMC) as a direct yaw moment controller in order to track the desired yaw rate and sideslip angle while also employing a joint distribution algorithm to control the torque distribution based on vehicle stability parameters, thereby enhancing system robustness; and the lower-level controller performs optimal torque allocation based on the optimal tire loading rate as the objective. A Speedgoat-CarSim hardware-in-the-loop simulation platform was established, and typical driving scenarios were simulated to assess the stability and accuracy of the proposed control algorithm. The results demonstrate that the proposed algorithm significantly enhances vehicle-handling stability across both high- and low-adhesion road conditions.

Optimization of In-Motion EV Charging Infrastructure for Power Systems Using Generative Adversarial Network-Based Distributionally Robust Techniques

This paper presents an innovative optimization framework for the co-management of dynamic electric vehicle (EV) charging lanes and power distribution networks, addressing grid stability amidst fluctuating EV charging demands. Integrating generative adversarial networks (GANs) and distributionally robust optimization (DRO), the framework models uncertainties in traffic flow and renewable energy generation, optimizing system performance under worst-case conditions to mitigate risks of grid instability. Applied to a highway with eight dynamic charging lanes (500 kW per lane), serving up to 50 EVs simultaneously, the framework balances energy contributions from 15 renewable generators (60% of the mix) and 10 non-renewable generators. Simulation results highlight its effectiveness, maintaining grid stability with voltage deviations within 0.02 p.u., reducing energy losses to under 0.8 MW during peak traffic (1500 vehicles per hour), and achieving 95% lane utilization. Dynamic charging enabled EV users to save USD 0.08 per kilometer through reduced stationary charging downtime, optimized travel efficiency, and lower energy costs. Additionally, the system minimizes maintenance costs by optimizing lane and grid reliability. This study underscores the potential of GAN-based DRO methodologies to enhance the efficiency of power grids supporting dynamic EV charging, offering scalable solutions for diverse regions and traffic scenarios.

Innovative on-off yaw dampers with switchable dynamic stiffness for enhanced stability of high-speed railway vehicles

The parameters of traditional yaw dampers are typically determined during the vehicle dynamics design stage based on specific wheel and rail profiles. However, dampers with fixed parameters often struggle to effectively mitigate hunting motions due to the significant variations in wheel-rail contact conditions encountered during operation. Unlike the traditional method of adjusting dynamic damping, the goal of this work is to propose an innovative on-off yaw damper that can realize switchable low and high dynamic stiffness, and thus has the ability to suppress both low-frequency carbody hunting and high-frequency bogie hunting under different wheel-rail conicity conditions. First, the working principle of the damper is detailed, highlighting switchable dynamic stiffness in two selectable modes. A physical model is then developed to study the damper’s dynamic characteristics under different combinations of excitation frequency and amplitude. The results reveal a substantial difference in dynamic stiffness between the two modes, with a 74% variance, while dynamic damping remains largely unaffected. A new parameter, termed dynamic stiffness difference, is introduced to describe the frequency-dependent stiffness property of the damper, and its sensitivity to damper parameters is explored, providing insights for design optimization. Finally, numerical simulations demonstrate that the on-off yaw damper significantly improves both low and high conicity stability, thereby enhancing overall vehicle stability across a wide range of wheel-rail contact conditions.

Based on AVL-CRUISE Hybrid and Electric Vehicle Simulation and FTP-72 Circle Power Consumption Factor Analysis

With the rapid advancement of technology and the growing scarcity of energy resources, finding effective ways to reduce power consumption has become increasingly crucial. Hybrid and electric vehicles play a vital role in this endeavor, warranting deeper exploration and focus. In this study, I utilized AVL-Cruise software, a powerful tool for simulation and data modeling, to conduct comprehensive simulations of hybrid and electric vehicle performance, with a specific focus on analyzing the FTP72 cycle.The simulation results are highly encouraging. The vehicle achieved a maximum speed of 164 km/h, an acceleration rate of 3 m/s, a 0 to 100 km/h acceleration time of 9 seconds, and a maximum climbing slope of approximately 74%. These figures not only demonstrate that hybrid and electric vehicles can meet performance expectations but also highlight their capability to deliver on key metrics. Furthermore, the total energy consumption was recorded at 4200 kJ, underscoring the efficiency of these vehicles.In addition to the simulation outcomes, I delved into various factors that influence vehicle performance, including vehicle mass, drag coefficient, and motor power.Using MATLAB, I conducted a detailed analysis of the relationships between these factors and energy consumption, incorporating residual analysis and deriving a mathematical equation to describe these interactions.Through this extensive data analysis and model fitting, we gain a deeper understanding of hybrid and electric vehicle dynamics. This research not only contributes to the field but also has the potential to drive further development, broadening the focus and scope of hybrid and electric vehicle technology