Precise battery remaining useful life (RUL) estimation is essential for battery system’s long-term management. Most current battery RUL estimation methods rely on historical data and make point estimations of RUL, which are incompetent when historical data is unavailable or a group of batteries is involved. To make battery RUL estimation independent of historical data and applicable for both a single cell and a group of cells, a life-decomposition-based battery RUL estimation framework is proposed in this paper. It decomposes RUL as the product of two random variables, extent of cycling (EOC) and whole life (WL), and forms RUL distribution based on current cycle’s data. For single-cell RUL estimation, the proposed method is tested on the open-access ISU-ILCC Battery Aging Dataset and achieves a relatively accurate estimation with a root mean square error (RMSE) of 306.13 cycles and an average normalized mean square percentage error (N-RMSPE) of 23.11 % on unseen battery cycling data. With more accurately observed WL information, these results are further improved to an RMSE of 217.77 cycles and an average N-RMSPE of 11.32 %. For battery group analysis, the RUL distribution is applied to the end-of-life warning from a group perspective. Warning threshold is calculated from RUL distribution’s mode and variance to indicate the existence of near-end-of-life cells, achieving an overall 0.8150 F2 score. Moreover, the predicted EOC may serve as a notable index for inconsistency among cells and battery sorting.

Range anxiety remains a major barrier to electric vehicle (EV) adoption, especially in charging-desert regions with limited infrastructure. Mobile charging stations (MCSs) offer a flexible on-demand solution, yet EV rescue planning in such regions has not been systematically investigated, particularly under velocity- and battery-state-of-charge-dependent energy constraints. This paper proposes an energy-constrained rescueplanning framework for an EV and a MCS on a road network. The framework integrates velocity-dependent energy modeling, reachable-subgraph construction, and a multi-objective rendezvous optimization that jointly accounts for energy consumption, travel time, and distance to the charging-desert boundary. A case study on the Kentucky–Tennessee boundary demonstrates that the proposed method can identify feasible meeting nodes and the globally optimal rescue strategy. Sensitivity analyses demonstrate the effects of state of charge reserve threshold, auxiliary load, and objective-function weights on feasibility and optimal speed selection.

This paper presents a model predictive control (MPC) approach for personalized lane-keeping automation that tracks driver-specific reference trajectories. Human subjects repeated runs across diverse driving conditions were conducted on a driving simulator. For each subject, the driving data are summarized into a representative trajectory that serves as the control reference and shows high-order shapes. The prediction model employs the Frenet-error dynamics and uses preview reference road curvature as a measured disturbance, ensuring consistent use of preview information. This control approach reduces phase lag and overshoot while consistently optimizing costs within constraints over the prediction horizon. In a dSPACE Automotive Simulation Models (ASM) environment, the proposed controller outperforms its baseline controller by improving tracking accuracy for multiple drivers in various driving scenarios, demonstrating the feasibility of personalized path tracking without additional feedforward loops and the practicality of the proposed framework.

Existing point-mass-based risk field models often fail to accurately quantify inter-vehicle risks, particularly in highly interactive maneuvers such as autonomous lane-changes. To address this issue, in this paper, we propose a Shape-Aware and Risk-Minimizing (SARM) trajectory planning framework for autonomous vehicle lane changes. We first explicitly consider polygonal vehicle shapes, and then define interaction risk as the closest-point-pair distance between vehicle shapes. The proposed shape-aware risk model is incorporated into lane-change trajectory planning via a bi-level optimization formulation. Next, the bi-level optimization problem is reformulated into a single-level Mathematical Program with Complementarity Constraints by introducing dual variables and analyzing the Karush–Kuhn–Tucker conditions, making the optimization problem tractable. Moreover, to enhance computational performance, a receding-horizon warm-start scheme is developed for continuous replanning. Simulations show that the proposed SARM planner achieves 26.42% and 16.75% lower risk compared with the point-mass-based method in two representative scenarios, while remaining computationally tractable for the receding-horizon lane-change planning.

Provides society information that may include news, reviews or technical notes that should be of interest to practitioners and researchers.

This paper presents a rule-based torque-split control strategy that minimizes real-time energy consumption for a dualmotor all-wheel-drive (AWD) electric vehicle (EV) based on the EcoCAR EV Challenge (EVC) platform. Assuming identical front and rear electric drive units (EDUs) and a convex approximation of motor losses, we show that the energy-optimal split lies into two operating modes: front-only at lower driver torque demand or equal split at higher demand. A speed-dependent torque demand defines the switching boundary between the two modes. The loss associated with the rear-axle decoupler when engaged is incorporated as a fixed term in the cost, showing the preference for front-only operation in low-load regions. The online implementation proceeds in two steps: (i) projection of the driver axle-torque request into the feasible set formed by EDU torque/current/power limits, battery current/power limits, brake-blending constraints, and chassis interventions, and (ii) selection between front-only and equal-split modes using a rule that accounts for the decoupler cost. To improve practical implementation and drivability, the two-mode rule is equipped with a mode-transition logic that includes explicit hysteresis and axle torque ramp limiting. This method avoids grid search while remaining interpretable and computationally efficient, suitable for embedded real-time control. The framework applies to both motoring and regenerative operations. The vehicle configuration and high-level requirements follow the EVC specification with utilization of the 2023 Cadillac Lyriq EV model.

Battery charging protocols play a pivotal role in the efficiency, safety, and longevity of battery systems. Existing protocol generation methods face three key limitations, including limited protocol diversity, inflexible across application scenarios, and lack of a compact unified representation that enables systematic analysis. To address these challenges, this article proposes a variational autoencoder (VAE)-based universal battery charging protocol (UBCP) generation framework that learns a unified, low-dimensional, latent representation capable of expressing charging protocols with comprehensive patterns. A randomized dataset is used to train the encoder, ensuring exposure to diverse patterns necessary for learning such a universal representation. A merge-based data augmentation strategy is introduced for the decoder to fuse distinct base protocols, improving reconstruction accuracy while preserving the ability to generate novel protocol patterns. The framework is evaluated in protocol reconstruction, fast-charging, and power-constrained charging scenarios. Experimental results show that the model achieves high-fidelity reconstruction and efficiently customizes protocols to meet given constraints. Compared with optimization-based baselines, the UBCP framework attains comparable or superior performance while requiring only a simple search and constraint check, highlighting its flexibility and computational efficiency.

High-speed path-tracking control for ground vehicles presents challenges that are exacerbated by inherent time delays in vehicle steering systems. Model Predictive Control (MPC) is widely used for path-tracking control due to its high accuracy and performance at high vehicle speeds. However, its complexity and high computational demand pose practical challenges. To address these issues, we propose a practical control design scheme, as an alternative, which integrates feedback, feedforward, and lead compensator components. The feedback component uses output feedback vehicle lateral errors from a look-ahead point, while the feedforward component provides curvature-dependent compensation. To enhance control bandwidth while preserving vehicle stability, we employed a lead compensator to construct the output feedback controller, optimized via a heuristic loop-shaping method. Simulation studies on a high-fidelity vehicle model were conducted to evaluate the proposed method and compared with an output feedback controller and an MPC in both lane-change and double-lane-change maneuvers. Compared to the baseline output feedback controller, the proposed approach achieved 52.98% and 48.61% reductions in root mean square error for vehicle lateral tracking in single and double lane-changes, respectively. Our method is comparable in performance and robustness to the MPC-based method for highway velocity range while requiring significantly lower computational effort.

Occluded areas restrict the field of view for autonomous vehicles, thus posing significant safety risks. Although cooperative perception via vehicle connectivity may mitigate these risks, gradual adoption of connected vehicles (CV) means such hazards will persist in mixed traffic in the foreseeable future. To enable safe autonomous overtaking in occluded zones under sparse CV deployment, this paper proposes a game-theoretical decision-making strategy using coordinated active perception. First, a risk assessment strategy based on reachability analysis quantifies maximum risk sets from potential occluded vehicles and derives safe drivable regions for the ego CV. Building on this, the coordinated active perception method employs automated CVs as active sensors that dynamically adjust their motion/pose to expand visibility and share information via wireless communications, thus minimizing overtaking conservativeness. Finally, with occluded vehicle intentions obtained via coordinated active perception, the ego CV interactively determines its maneuver based on a potential differential game, which is transformed into a centralized optimal control problem and solved using an iterative Linear Quadratic Regulator. Simulation results demonstrate the potential enhancement of overtaking efficiency through coordinated active perception, while the proposed interaction strategy ensures robust safety and adaptability across diverse overtaking scenarios.

Accurate remaining useful life (RUL) prediction is crucial to long-term battery management. Existing RUL prediction methods have limited applicability due to their reliance on historical data and the prerequisite of future working condition being the same as the past. A historical-data-independent RUL prediction method that only needs current cycle’s data is proposed in this paper. By a life decomposition method, battery RUL is separated into three elements, including mean whole life, extent of cycling, and deviation, which serve as the ingredients for the RUL calculation. Compared with models that predict RUL directly, the proposed method outperforms its counterparts and achieves the best performance on the test set. Additionally, when utilizing the prior knowledge of the number of past cycles, the proposed method achieves comparable prediction performance with the state-of-the-art, historical-data-dependent methods. This work sheds light on more practical battery RUL estimation and offers a unique life decomposition method that is beneficial to battery RUL prediction and analysis.

The uncertainty in human driving behaviors leads to stop-and-go traffic congestion on freeway. The freeway traffic dynamics are governed by the Aw-Rascle-Zhang (ARZ) traffic Partial Differential Equation (PDE) models with unknown relaxation time. Motivated by the adaptive traffic control problem, this paper presents a neural operator (NO) based adaptive boundary control design for the coupled 2 x 2 hyperbolic systems with uncertain spatially varying in-domain coefficients and boundary parameter. In traditional adaptive control for PDEs, solving backstepping kernel online can be computationally intensive, as it updates the estimation of coefficients at each time step. To address this challenge, we use operator learning, i.e. DeepONet, to learn the mapping from system parameters to the kernels functions. DeepONet, a class of deep neural networks designed for approximating operators, has shown strong potential for approximating PDE backstepping designs in recent studies. Unlike previous works that focus on approximating single kernel equation associated with the scalar PDE system, we extend this framework to approximate PDE kernels for a class of the first-order coupled 2 x 2 hyperbolic kernel equations. Our approach demonstrates that DeepONet is nearly two orders of magnitude faster than traditional PDE solvers for generating kernel functions, while maintaining a loss on the order of 10(-3). In addition, we rigorously establish the system's stability via Lyapunov analysis when employing DeepONet-approximated kernels in the adaptive controller. The proposed adaptive control is compared with reinforcement learning (RL) methods. Our approach guarantees stability and does not rely on initial values, which is essential for rapidly changing traffic scenarios. This is the first time this operator learning framework has been applied to the adaptive control of the ARZ traffic model, significantly enhancing the real-time applicability of this design framework for mitigating traffic congestion. (c) 2025 Elsevier Ltd. All rights are reserved, including those for text and data mining, AI training, and similar technologies.

This paper presents a personalizable, physics-based model that quantifies individual drivers' expected attentiveness under varying driving conditions in vehicle lane-keeping automation. The model introduces a physically interpretable formulation for drivers' cognitive load as a function of vehicle speed and road curvature. It leverages intuitive, personalized indicators derived from gaze data, offering greater interpretability than conventional gaze metrics and enabling driver-specific customization. Using a high-fidelity driving simulator and an eye-tracking system, we collected objective gaze data and applied a hybrid method combining subjective ratings with the NASA Task Load Index to optimize model parameters. We evaluated the model's predictive performance and reliability with human subject experiments from multiple perspectives. This model supports human-centric vehicle automation by estimating if a driver is under-, over-, or appropriately attentive.

Accurate gross mass estimation is critical in operations of electric vehicles that experience payload variations due to its predominant impact on energy consumption and dynamics. This paper introduces a novel Immersion and Invariance (I&I) adaptive observer for estimating gross vehicle mass using only vehicle velocity and driving torque signals. Compared to conventional adaptive observer designs, our approach offers two distinct advantages. First, the I&I scheme is "no regret", ensuring that the norm of parameter estimation errors remains non-expansive. Second, the design enables asymptotic recovery of deterministic observer error dynamics, as the I&I adaptation drives parameter-error-induced perturbations toward an attractive and zeroing manifold. Additionally, an observer incorporating drivetrain inertia is proposed to account for rotational dynamics and enhance model fidelity. The effectiveness of the observers is demonstrated through high-fidelity, hardware-in-the-loop, experiments.

Visual displays are used in vehicle automation systems to communicate the vehicle's perception of the surrounding environment to the driver and passengers. This visual communication may impact how humans interact with the vehicle during driving. Thus, careful design of vehicle automation system visual communication is important for vehicle-driver collaboration. However, there is a lack of systematic study on how the level of detail in vehicle automation visual communication affects human driver's workload, engagement, and acceptance. This paper presents a pilot research aiming to assess the impact of visual communication level of detail in vehicle automation systems. Both objective evaluation and subjective evaluation are conducted with human driving data collected on a driving simulator. Experimental results show a good agreement between the proposed objective assessment and subjective assessment.

In mixed traffic scenarios where human-driven vehicles (HDVs) and autonomous/automated vehicles (AVs) of various levels coexist, the lateral behaviors of all vehicles such as lane-change maneuvers are influenced by surrounding vehicles. Predicting these behaviors before the lane-change maneuvers is essential for improving the decision-making and path-planning strategies of AVs, while offering valuable driving references for human drivers. This paper proposes a universal lateral motion prediction model applicable for both HDVs and AVs in mixed highway traffic. First, variables at the onset of the lane change are analyzed, and their respective relationships with lane-change time, peak lateral speed, and the time to reach peak lateral speed are established. Next, a mapping function is designed to preprocess raw variables (e.g., time to collision) and connect it with the proposed prediction model for enhanced interpretability. To exemplify the predictability of the model, two drivers were recruited for repeated lane change tests in diverse scenarios, generating raw data for their model parameter estimation. Experimental results demonstrate that the proposed model provides reliable lateral motion prediction for vehicles under various scenarios and drivers, establishing a foundation for enhancing decision-making and path-planning in AVs within mixed traffic environments. Copyright (c) 2025 The Authors.

Repeated driving along the same route is common in the real world and promotes familiarity, which significantly influences driving behavior and safety. While prior studies have characterized familiarity using aggregated route-level trends, little attention has been given to how it manifests across diverse scenarios and individuals. This study adopts a counterexample-driven approach to examine whether such global summaries adequately reflect nuanced behavioral adaptations across contexts and drivers. A preliminary study was conducted using a driving simulator with eye tracking. Participants repeatedly drove a virtual urban route containing intersections, curves, and regulatory signs. Gaze behavior was analyzed using stationary and dynamic metrics to capture changes in visual attention. Results show that while global route-level trends indicate increased driving speed and decreased gaze entropy, scenario- and individual-level analyses reveal substantial variability. Notably, the common assumption that fixation duration on traffic signs uniformly decreases with increasing familiarity is challenged, as one subject continued using stop signs as visual references for stopping. These findings highlight the limitations of relying solely on global analyses and underscore the value of incorporating scenario- and individual-level perspectives to better understand how driving familiarity develops. This shift may inform the design of adaptive, human-centric autonomous systems with improved safety and efficiency.

The uncertainty in human driving behaviors leads to stop-and-go instabilities in freeway traffic. The traffic dynamics are typically modeled by the Aw-Rascle-Zhang (ARZ) Partial Differential Equation (PDE) models, in which the relaxation time parameter is usually unknown or hard to calibrate. This paper proposes an adaptive boundary control design based on neural operators (NO) for the ARZ PDE systems. In adaptive control, solving the backstepping kernel PDEs online requires significant computational resources at each timestep to update estimates of the unknown system parameters. To address this, we employ DeepONet to efficiently map model parameters to kernel functions. Simulations show that DeepONet generates kernel solutions nearly two orders of magnitude faster than traditional solvers while maintaining a loss on the order of 10-2. Lyapunov analysis further validates the stability of the system when using DeepONet-approximated kernels in the adaptive controller. This result suggests that neural operators can significantly accelerate the acquisition of adaptive controllers for traffic control. Copyright (c) 2025 The Authors. This is an open access article under the CC BY-NC-ND license (https://creativecommons.org/licenses/by-nc-nd/4.0/)