Browse Topic: Autonomous vehicles

Items (3,365)
Find
Trajectory Optimization for Non-Cooperative Target Spacecraft Based on Sequential Convex Programming2026-99-2330To be published on 10/30/2026
Autonomous operations near space targets that do not cooperate present great challenges because position and attitude planning must be linked. A target of this type lacks navigation aids or communication systems, which makes relative motion planning much more difficult. Although methods based on convex optimization work well for problems with three degrees of freedom, like planetary landing, their use in integrated operations with six degrees of freedom is limited by constraints that are not convex. To solve this problem, this paper proposes a new framework for servicing spacecraft using sequential convex programming. The primary contribution is the extension of convex methods to high dimensional trajectory planning while adding a robust trust region constraint based on trajectory deviation. This mechanism reduces the chance of finding no solution during convex approximations. It improves reliability and convergence when solving complex problems. The approach first creates a continuous optimal control problem with an index that balances energy use and the angle of the line of sight. Nonlinear dynamics are simplified using first order Taylor expansions. Constraints for the field of view and collision avoidance are modeled using spherical and ellipsoidal regions. The non convex problem is changed into a sequence of discrete convex subproblems and solved through iterations. Numerical simulations show that the servicing spacecraft can safely approach the target with stable attitude while meeting all physical limits. The method using sequential convex programming with a trust region demonstrates great potential in optimizing computational efficiency alongside strong numerical stability, laying a solid foundation for future real-time autonomous trajectory generation in complex orbital environments.
Wang, Yuzhi, Yao, Zihao, Wang, Qian, Han, Ziwen, Zhang, Chao, Zhang, Yumei, Li, Shunli
Configuration Design and Implementation of the SuperView Neo 2 Satellite System2026-99-2370To be published on 10/30/2026
With the rapid advancement of spaceborne Synthetic Aperture Radar (SAR) technology, distributed Interferometric SAR (InSAR) has emerged as the definitive method for high-precision global topographic mapping. This study focuses on the SuperView Neo 2 (SVN2) satellite system, developed by China Siwei Surveying and Mapping Technology Co., Ltd., detailing its technical configuration design and operational implementation within the next-generation commercial remote sensing sector. Equipped with X-band high-performance phased-array SAR payloads, the SVN2 system is engineered for the high-efficiency production of global 1:25,000-scale Digital Surface Models (DSM) while providing millimeter-level surface deformation monitoring through Differential InSAR (DInSAR) capabilities. The paper emphasizes a systematic design approach that integrates four coupled elements: mission objectives, system performance, orbit design, and formation geometry. To satisfy the rigorous requirements of interferometric mapping, the system adopts a sun-synchronous, strictly repeating frozen orbit with a 200-meter radius “pipeline” control strategy. By implementing flexible formation configurations for high and low latitudes, SVN2 achieves the triple synchronization of time, space, and phase necessary for sub-meter resolution and high vertical accuracy. Using the SVN2-03/04 cluster as a primary case study, the results demonstrate that through a sequence of over 40 precise orbital maneuvers, the dual-satellite system successfully established a close-range fly-around formation within the 100-meter-level pipeline just seven days after launch. This operational success represents three major technical breakthroughs for Chinese commercial radar satellites: high-precision strictly repeating orbit maintenance, stable close-range formation flying, and fully autonomous inter-satellite cooperative control. The SVN2 constellation provides a robust, all-weather mapping guarantee for the global geographic information framework and supports diverse applications in geological hazard detection and social economic planning.
Tang, Ke, Qiao, Yishi, Huang, Tao, Zhang, Guoliang, Zhao, Di, Zhang, Zhongze
Suppression of Traffic Oscillations by Vehicular Platoons with Periodic Control Configuration12-10-01-000310/21/2026
Traffic oscillations degrade the operational efficiency and safety of vehicular platoons. This study proposes a predecessor-following platoon with periodically assigned feedback gains and investigates its frequency-selective disturbance attenuation mechanism. An auxiliary second-order undamped periodic model is first established, and the Bloch–Floquet formulation is used to derive the nominal dispersion relation and bandgap boundaries. Closed-loop disturbance attenuation is then evaluated separately using spacing-error transfer matrices that retain positive velocity feedback and actuator-response lag. Frequency-domain transmission analysis and time-domain simulations show that, under nominal homogeneous and fully controllable connected and automated vehicle (CAV) assumptions, the periodic gain configuration creates tunable attenuation bands and provides stronger downstream attenuation than a homogeneous-gain configuration. The results clarify the relationship among periodic gain assignment, bandgap formation, periodic-cell string stability, and finite-length traffic-wave suppression. The effects of vehicle heterogeneity, sensing delay, and mixed traffic remain topics for future investigation.
Yang, Xiujian, Fan, Yanwen, Zhuang, Qingyuan, Qiao, Feiyu
Vehicle Attitude Control via Load Resistance Regulation of a Hydraulic–Electric Energy-Harvesting Suspension10-11-01-000710/21/2026
Vehicle dynamics during autonomous driving can cause passenger discomfort as well as motion sickness. This article introduces an energy-efficient vehicle attitude control method through load resistance control in a hydraulic–electric energy-harvesting suspension. A thorough study on how external load resistance affects a vehicle’s pitch, roll, and vertical acceleration has been conducted without requiring additional energy input. To analyze the effects that the variation of load resistance has on vehicle attitude, a 10 DOF vehicle model and a hydraulic–electric energy-harvesting suspension have been developed. The relationship between load resistance and vehicle attitude is defined and experimentally demonstrated in slalom tests under real-world conditions. The results demonstrate that reducing the load resistance from 70 Ω to 30 Ω reduces peak roll angle by 33.9% and peak pitch angle by 30.4%, while the RMS values are reduced by 46.2%, 28.6%, and 28.6% for roll angle, pitch angle, and vertical acceleration, respectively. A frequency-domain analysis shows that these improvements are achieved without shifting vibrational energy into the 4–8 Hz range most perceptible to humans. This systematic approach provides an energy-efficient way to improve vehicle stability and ride comfort, with potential implications for reducing motion sickness in autonomous driving systems. The quantitative relationship between load resistance and vehicle attitude provides a new understanding of the dynamics of a vehicle with energy-harvesting suspension systems.
Fan, Zhengwu, Liu, Wenbo, He, Mengfan, Xu, Guangzhao, Wu, Zhifei
2026 International Symposium on Aerospace Information and Industry Innovation2026-99-290010/19/2026
TOC
Tobolski, Sue
Satellite Autonomous Task Decision-Making Method Based on Attention Pointer Network2026-99-296510/19/2026
Large-scale satellite constellation on-orbit missions are characterized by multiple objectives, multiple constraints, and strong dynamism. Traditional ground control and management modes have become difficult to meet the requirements for autonomous real-time decision-making by satellites. This paper addresses the need for autonomous and rapid decision-making capabilities on-board satellites, and proposes a satellite autonomous task planning and decision-making method based on an attention pointer network, addressing issues such as limited on-orbit computing resources and insufficient generalization of planning methods. Firstly, by analyzing and mathematically describing the characteristics and constraints of satellite tasks, a Markov model for satellite autonomous task planning problems is constructed. Secondly, the pointer network is combined with a reinforcement learning framework, and an attention mechanism is introduced to perform probabilistic learning on the execution selection of satellite tasks, generating an optimized sequence of satellite task planning results. Finally, simulation experiments show that after training, this method exhibits certain generalization capabilities and significantly superior computational efficiency compared to traditional intelligent optimization algorithms such as genetic algorithms. It can adapt to the single-satellite autonomous dynamic planning requirements under large-scale constellation control and management, providing technical support for autonomous operation and online intelligent task planning of satellites in orbit.
Feng, Xiaoen, Li, Yuqing, Qiu, Jianbin, Li, Meicheng
Task Allocation and Trajectory Optimization for Mother-Daughter Satellite Clusters Observing Non-Cooperative Constellations via Improved Genetic Algorithm2026-99-291710/19/2026
Over the past decade, the rapid deployment of Low Earth Orbit (LEO) mega-constellations has significantly reshaped the spatial distribution of the near-Earth environment and increased orbital density to historically high levels. While popularizing space access, the accumulation of defunct or non-cooperative satellites raises severe concerns regarding space sustainability and the Kessler syndrome. Consequently, the ability to conduct active, multi-target close-proximity inspection is becoming an essential capability for Space Situational Awareness (SSA). Traditional monolithic servicing strategies rely on sequential point-to-point orbital transfers, which have to face heavy computation cost and restrictive fuel constraints when targeting multi-planar distributed networks requiring continuous out-of-plane maneuvers and phasing adjustments. In order to overcome these limitations, the paper proposes a coordinated mother-daughter satellite cluster architecture. A primary mother satellite serves as a deployment hub, while a swarm of highly maneuverable daughter satellites executes sequential "Observe-and-Return" missions, gathering high-fidelity SSA data. Assigning these multi-target inspection missions to a limited fleet creates an NP-hard combinatorial challenge, further compounded by strictly non-overlapping operational time windows and rigid propellant boundaries. To navigate this high-dimensional search space, the study introduces an improved Genetic Algorithm (GA) featuring a dynamic chronological decoding mechanism. Furthermore, to eliminate secular positional drifts during orbital integration, a high-fidelity five-impulse maneuver sequence is established, embedding an autonomous Lambert-based positional compensation mechanism. Simulation results indicate that the proposed method yields scalable, fuel-feasible, and load-balanced kinematic baselines for autonomous target characterization across large-scale constellations. Additionally, an event-triggered dynamic re-planning mechanism is utilized to robustly adapt to uncooperative evasive maneuvers in highly dynamic real-world space scenarios.
Sheng, Beining, Tian, Yang, Li, Tieying, Shan, Yiming
A Wide-Angle Beam Scanning Algorithm for Reconfigurable Liquid Crystal Phased Arrays in Low Earth Orbit Satellite Acquisition2026-99-294910/19/2026
To satisfy the real-time beam tracking requirement of low earth orbit (LEO) satellite ground terminals under highly dynamic pass conditions, this paper proposes a three-state hierarchical state-machine beam-control algorithm for liquid crystal phased array antennas (LCPAs). The tracking process is divided into three stages. S1 coarse nine-point acquisition is used to rapidly cover a wide pointing-uncertainty region. S2 fine nine-point refinement introduces quadratic-surface fitting to obtain local peak estimation at sub-step resolution. S3 predictive tracking adopts an interacting multiple model Kalman filter (IMM-KF) based on the constant-acceleration (CA) model and the Singer model, and designs an adaptive covariance regulation mechanism for both measurement and process noise. Combined with explicit delay compensation for liquid-crystal response time, the method suppresses pointing lag and error accumulation in high-dynamic intervals. A dual-threshold hysteretic loss-of-lock criterion and a filter-reset mechanism are also integrated to guarantee autonomous link recovery. Simulation results show that, in an LEO pass scenario with a maximum angular rate of about 2 °/s, the proposed method reduces the median 2D RMSE by about 52%, reduces the 90% quantile error by about 54%, and lowers the median average power loss from 0.14 dB to 0.03 dB compared with the traditional scheme.
Tian, Lingrui, Guo, Xiaoxu, Xu, Yao, Chen, An, Zhu, Yaowei, Zhang, Zhizhong
EAU-Net: An Edge-Aware Enhanced Attention U-Net for Robust SAR Near-Shore Segmentation2026-99-291010/19/2026
Synthetic Aperture Radar (SAR) provides important all-weather images. These images help researchers observe the sea surface. However, separating the coast from the ocean in these images is hard. Near-shore areas face three main issues. The first issue is coherent speckle noise. The second issue is geometric distortion caused by the radar’s side-looking geometry. The third issue is feature aliasing caused by multiple scattering from coastal buildings. To solve these problems, this paper proposes the Enhanced Attention U-Net (EAU-Net). This model is a deep learning framework. It adds SAR scattering priors directly into the learning process. The architecture has three main parts. First, the encoder includes an adaptive local signal-to-noise ratio (SNR) module. This module filters multiplicative speckle dynamically. It stops the network from mistaking noise for real features. Second, the network bottleneck uses a lightweight Transformer block. This block learns global shape priors. It helps the network find the continuous coastline even when radar shadows cover parts of the image. Third, the decoder uses an adaptive edge refinement mechanism. A composite edge-aware loss guides this mechanism. It fixes rough edges and corrects pixel boundaries. This step removes jagged artifacts. Tests on a near-shore dataset show the reliability of this method. EAU-Net achieves a Dice Similarity Coefficient (DSC) of 0.9601 and an Intersection over Union (IoU) of 0.9234. The Average Symmetric Surface Distance (ASSD) is 5.7501 pixels. This paper compares EAU-Net with classical convolutional networks, global-attention architectures, and vision foundation models. The results show that EAU-Net preserves boundary precision better. It outperforms state-space models like VM-UNet and large-scale zero-shot models like SAM-Med2D. EAU-Net uses physical inductive biases to separate coastal building interference from marine targets. This provides a realistic foundation for downstream tasks like autonomous ship detection.
Bai, YuanJie, Zhong, Sheng
A Genetic Algorithm-Based Autonomous Mission Planning Method for Limited-Thrust Modes2026-99-294810/19/2026
To meet the requirements of rapidity, economy and efficiency for space debris removal missions and space debris governance missions, this study designs an autonomous mission planning scheme under finite thrust conditions, to achieve the disposal of multiple target space debris within a single orbital period. The scheme decomposes the planning problem into two sub-problems: mission object sequence optimization and transfer orbit design, and establishes mathematical models from the perspectives of global mission scheduling and local trajectory optimization respectively. An improved genetic algorithm is adopted for solution, to ensure the optimization of resource utilization under fuel constraints. This study systematically elaborates the theoretical basis, optimization indicators and constraint modes of the planning method, constructs a complete technical framework, and verifies the correctness and effectiveness of the method through numerical simulations, providing methodological guidance for the autonomous implementation of in-orbit rendezvous missions.
Yu, Dateng, Guo, Peng, Li, Ziyi
A CTCM-ALA Adaptive Embedded Fusion Optimization Method for UAV Path Planning in Complex Medical Rescue Environments2026-99-290510/19/2026
To improve the accuracy and safety margin deficiencies of unmanned aerial vehicles in autonomous medical emergency search-and-rescue missions, this paper develops an adaptive embedded fusion optimizer named CTCM-ALA for rescue UAV path planning. The proposed algorithm hybridizes the tribe-competition member-cooperation strategy and artificial lemming algorithm to construct hierarchically adaptive initial populations, speed up convergence and prevent local convergence traps. Tests on 12 CEC2017 functions prove its better optimization performance. Real-world applications in dense urban and complicated topographic medical rescue cases verify its reliability and precision.
Chen, Yuhan, Xiong, Junhui, Li, Yuesheng, Zhou, Yifan, Yao, Junjie, Li, Rujun
Gradient-Driven Point Allocation for Depth-Guided 3D Gaussian Splatting in Indoor Scenes2026-99-299410/19/2026
High-precision 3D environmental perception is crucial for autonomous systems to navigate and interact effectively within complex indoor environments. However, achieving high-granularity 3D reconstruction and high-fidelity rendering remains challenging. Although 3D Gaussian Splatting (3DGS) enables efficient photorealistic rendering, its geometric inconsistency limits its ability to reconstruct high-accuracy scenes. Standard 3DGS relies on sparse Structure-from-Motion (SfM) initialization, which often leads to uneven point distributions and floaters, particularly in weak-texture areas like plain walls. Furthermore, existing methods based on depth regularization or geometric constraints usually fail to address the fundamental sparsity of the initialization. To address these issues, we propose a lightweight gradient-driven point allocation strategy for depth-guided Gaussian splatting that bypasses traditional SfM. By utilizing RGB-D priors, we obtain a high-quality, structurally continuous initial Gaussian distribution through multi-view depth back-projection and texture-aware density control. Our method incorporates a bidirectional re-projection check to eliminate transient noise and occlusions, ensuring geometric consistency. We introduce a dual-grid allocation scheme that adaptively distributes Gaussians based on spatial frequency, mandating density in high-gradient regions while enforcing sparsity in flat areas. Quantitative and qualitative experiments on the ScanNet++ dataset demonstrate that our method achieves superior reconstruction accuracy and rendering quality. Notably, our approach achieves state-of-the-art performance with a significantly reduced Gaussian count, decreasing from 1.70M to 0.39M in representative scenes, thereby mitigating unnecessary point cloud bloat and reducing memory overhead without sacrificing structural integrity.
Shang, Yuhao, Qu, Zhenshen, Rong, Guanyi, Gao, Qiushuang, Yu, Zhiwei, Bakar, Elmi Bin Abu, Junos, Mohammad Haniff Bin
A Dynamic Primary/Backup Path Algorithm Based on Multi-Level Feedback Queue for Differentiated Services in LEO Satellite Networks2026-99-297710/19/2026
With the accelerated deployment of Low Earth Orbit (LEO) mega-constellations, such as Starlink and OneWeb, satellite networks are transitioning from providing best-effort connectivity to supporting stringent, mission-critical services, including autonomous driving coordination, ultra-high-definition video streaming, and real-time financial transactions. This paradigm shift imposes highly heterogeneous Quality of Service (QoS) requirements regarding latency, bandwidth, and reliability. However, traditional unified scheduling and static routing mechanisms are inadequate for handling the dynamic topologies and bursty traffic inherent in LEO environments, often leading to service starvation or excessive packet loss. To address these challenges, this paper proposes an integrated primary-backup path guarantee framework that tightly couples a Multi-Level Feedback Queue (MLFQ)-based dynamic scheduling algorithm with a Dynamic Weight Adaptive Routing Algorithm (DWARA). The MLFQ scheduler employs an aging-based priority promotion mechanism to dynamically adjust time slices, effectively preventing low-priority traffic starvation while ensuring ultra-low latency for Class A services. Concurrently, DWARA models QoS as a three-dimensional vector (Throughput, Delay, Reliability) and utilizes the eigenvector method to derive service-aware weights for path selection. To validate the efficacy of the proposed scheme, comprehensive simulations were conducted using STK and Python, emulating a 64-node polar constellation under varying fault conditions. The results demonstrate that the integrated framework significantly outperforms baseline algorithms. Specifically, it reduces the average blocking rate by up to 37% under scenarios with 10 faulty satellites compared to Dijkstra, while lowering end-to-end latency for delay-sensitive services to approximately 28 ms at a traffic arrival rate of 500. Furthermore, the framework achieves a 92% bandwidth satisfaction rate for high-throughput services and maintains near-zero packet loss for high-reliability transmissions, offering a scalable and robust QoS assurance solution for next-generation space-air-ground integrated networks.
Zhang, Hongyuan, Zhang, Pengfei, Zhou, Qihao, Gao, Zhao
SAE Truck & Off-Highway Engineering: October 202626TOFHP1010/15/2026
MAHLE axes range anxiety for commercial EVs New onboard range extender aims to accelerate the electrification of heavy-duty trucks thanks to less weight and reduced TCO. Ground defense goes driverless A new generation of military ground vehicles has arrived with robotic and fully autonomous capability thanks to recent advances in AI and other technologies. Cummins delivers on a lighter Class 8 driveline How the new RPL35+ was engineered to reduce weight without conceding on any other factors - and why the hollow spline sleeve at its core took three years to get right. Editorial Defense goes on advanced-tech offensive Self-driving vehicles demand new cybersecurity thinking Hydrogen ICE development takes big steps toward commercialization Cost savings, circular economy benefits core to steady growth for heavy-duty engine reman John Deere launches JD, an AI assistant to enhance farmers' efficiency and eventually assist operators in the cab Product Briefs SPOTLIGHTS: Connectors; Hydraulics and Flow Control Q&A Westport's dual fuel injector enables sparkless CNG ignition
Latency-Aware Optimal Scheduling Scheme for Computational Offloading12-10-01-000210/9/2026
Proactive safety and vehicle automation typically requires high energy use from sensors and energy-intensive computing from sensor data processing. For high-quality, reliable perception and localization within a driving environment at the vehicle level, incoming data from multiple sensors need to be fused using advanced computational algorithms, which demand a high compute load. Alternatively, computational offloading of automated driving tasks shifts energy consumption from the vehicle to cloud infrastructure, where renewable energy sources, such as hydropower or solar power, can be utilized more efficiently. Herein, an optimal scheduling strategy for autonomous driving tasks via the cloud layer is formulated as a mixed-integer linear programming (MILP) problem and verified using measurement-informed task graphs. It is shown that cloud-based computational offloading enables energy-efficient operation while maintaining task deadlines to ensure the timely availability of perception and localization information, which is consistent with prior studies in the literature. Simulation results demonstrate that on average, 35.78% of the computational load was offloaded to the cloud, which can achieve significant energy savings for the onboard system. The compute operations achieved an average energy savings of 35.65%, while total system savings ranged from 26.16% to 30.67% under different cloud energy efficiency scenarios, highlighting the advantages of offloading compute-intensive tasks. The framework was further evaluated using multiple directed acyclic graph (DAG) configurations to assess its scalability and adaptability. Critical tasks, such as sensor fusion, were executed exclusively on the vehicle to ensure real-time responsiveness.
Sharma, Sachin, Meyer, Richard, Feinberg, Ben, Asher, Zachary
Mobilizing Innovation: Venture Capital Alignment for Defense with an Autonomous Construction Case Study2026-01-75609/22/2026
Defense acquisition often struggles to match the pace of private investment, slowing the transition of mature commercial technologies into military use. This paper examines how aligning government acquisition with venture-oriented business models can increase industry participation, accelerate fielding, and reduce government program office risk. Using autonomous construction as a case study, it highlights how commercial investment has advanced autonomy while traditional procurement limits adoption. The paper outlines approaches such as non-traditional partnerships, phased acquisition, and performance-linked revenue structures to improve flexibility, leverage private capital, and expand the Defense Industrial Base while speeding operational capability delivery. Citation: Mazzara, M., San Nicolas, A., Gadea, J., Himmel, M., Kruger, J., Gill, C., & Simon, A., Soylemezoglu, A., Netchaev, A., Nottage, D., Klein, J. “Mobilizing Innovation: Venture Capital Alignment for Defense with Autonomous Construction Case Study” In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA Michigan Chapter, Novi, MI, August 11–13, 2026.
Mazzara, Mark, Nicolas, Austen San, Gadea, James, Himmel, Max, Kruger, John, Gill, Charles “Spuck”, Simon, Andrea, Soylemezoglu, Ahmet, Netchaev, Anton, Nottage, Dustin, Klein, Jordan
Physics-Constrained Procedural Generation of Synthetic Environments for the Robust Validation of Autonomous Maneuver and Logistics2026-01-75499/22/2026
The validation of Autonomous Ground Vehicles (AGVs) and intelligent logistics planners is frequently compromised by the ”Sim-to-Real” gap, where simulation environments fail to replicate the physical friction of operational deployment. Ideally, valid test cases must enforce strict mobility constraints and impose realistic sustainment penalties; however, many current generation tools rely on idealized terrain interactions and infinite-resource assumptions. We present a real-time procedural framework designed to generate high-friction validation environments that stress-test the robustness of the System Under Test (SUT). The architecture integrates gradient-based terrain analysis with a stochastic contested logistics model. It ingests synthetic heightmaps to precompute mobility corridors, ensuring that every generated evaluation episode adheres to vehicle-specific traversability limits. Simultaneously, a logistics kernel enforces fuel consumption scaled by terrain gradients and models supply chain interdiction as a parameterized Bernoulli process. We validate this framework through a ”Digital Twin” methodology, demonstrating that terrain-aware generation eliminates invalid initialization states (0% mobility violations) while the logistics model induces operationally relevant failure modes in the SUT. This unclassified, open-architecture approach supports DoD Verification, Validation, and Accreditation (VV&A) requirements by providing deterministic, reproducible edge cases for autonomous system evaluation.
Soykan, Bulent, Rabadi, Ghaith, Bochenek, Grace, Paul, Victor J.
Multimodal Digital-Twin Simulations for Off-Road Autonomous Vehicles2026-01-75509/22/2026
Research in autonomous driving has largely focused on structured, on-road environments in densely populated urban areas, leaving off-road autonomy comparatively underexplored despite its importance for applications such as search and rescue, agriculture, and defense. Progress in developing solutions for off-road autonomy is heavily constrained by the high cost, safety risks, and limited coverage of real-world off-road data collection, particularly for rare terrain-induced failure cases. To address these challenges, we introduce a large-scale simulation platform for off-road autonomous driving generated in the Unity3D engine. Our simulator provides indefinite data from multiple RGB cameras, LiDAR, GPS, and IMU sensors to support research in long-duration applications. Our platform also facilitates research in risk-aware planning through dynamic weather, lighting, and annotations for sequence-level failures like collisions and weather-induced sensor blockages. By combining long-duration temporal coverage, multimodal sensing, and procedurally generated terrain diversity, our simulator facilitates progress for systematic evaluation of learning-based autonomous driving models in unstructured, safety-critical environments.
Ross, Timothy, Boone, Julia, Afghah, Fatemeh
Clemson Off-Road Dataset (CORD): A Multimodal Benchmark for Autonomy beyond the Flat-World2026-01-75189/22/2026
While autonomous perception has matured within the structured confines of urban roadways, it remains brittle when confronting the chaotic, non-rigid terrain of the natural world. This paper introduces the Clemson Off-Road Dataset, a high-fidelity, multimodal dataset engineered to bridge this gap by challenging standard “flat-world” assumptions. Featuring 2.90 TB of sensor data, the dataset captures a diverse spectrum of unstructured environments, ranging from the transitional trails of CU-ICAR and the day/night lighting dynamics of TN3 to the unstructured wilderness of Camp Daniels and the novel coastal scenery of Edisto Island. Distinguishing itself from existing forest-centric benchmarks, the Clemson Dataset provides a first-of-its-kind focus on coastal data, featuring unique adversarial conditions such as extreme solar glare, loose sand, and shifting tide lines. The data is collected aboard a Polaris RZR Pro R 4, a high-performance platform integrated with a sensor suite designed to perceive physics beyond geometry. Alongside 360° HD camera coverage, 3D LiDAR, and Radar, we integrate Cubert Ultris Hyperspectral imaging and Prophesee EVK4 Event-based vision to enable material classification and high-dynamic-range motion tracking. To overcome the bottleneck in ground truth generation, we used our “AI LabelMate,” a context-aware semi-automated annotation agent that fuses Vision-Language Models (Florence-2) with SAM2 to generate 6331 pixel-perfect annotated frames using a specialized off-road ontology and a human-in-the-loop pipeline. We establish performance baselines using Oneformer for semantic segmentation and used SalsaNext for lidar point clouds labelling. Available in both raw ROS2 bag and extracted standard formats, this Dataset serves as a pivotal testing ground for the next generation of robust autonomous systems.The dataset of this paper is available upon request to the Virtual Prototyping of Autonomy-Enabled Ground Systems (VIPR-GS) Center.
Patil, Ashish, Gupta, Prakhar, Bhosale, Mayuresh, Mukwaya, Arthur, Jegede, Akinbobola, Mikulski, Dariusz, Mwakalonge, Judith, Jia, Yunyi
Hybridnav: A Hybrid Approach to Zero-Shot Object Detection with Adaptive Exploration for Autonomous Reconnaissance2026-01-75179/22/2026
Autonomous reconnaissance in unknown or contested environments demands robust perception systems capable of identifying diverse objects without prior training data. This paper presents HybridNAV, a hybrid framework that combines multiple foundation models with an adaptive navigation system for zero-shot object detection and autonomous exploration. Unlike monolithic detection models, HybridNAV’s multi-model fusion achieves balanced precision (0.60) and recall (0.58) with a macro F-score of 0.59, representing a 24% improvement over single-model baselines. The adaptive navigation system reduces scan time by 25% and path length by 30% compared to static waypoint approaches, while operating in real-time at 3.2 Hz with sub-100 msec latency on resource-constrained hardware. All processing is performed locally on the robotic platform, eliminating reliance on external communication infrastructure—a critical requirement for operations in communication-denied environments. We evaluate HybridNAV in both simulated indoor scenes and physical robot trials, demonstrating its effectiveness for intelligence gathering in unknown environments.
Indurthi, Hemanth, Martinson, Eric
Bullfrog: AI Powered Counter Unmanned Aerial System2026-01-75199/22/2026
The proliferation of small unmanned aircraft systems (sUAS) presents an asymmetric threat to ground maneuver forces operating in contested and gray-zone environments. The Bullfrog Autonomous Weapon Station (AWS) addresses this operational gap through a passive, AI-powered counter-UAS system employing computer vision and machine learning for autonomous detection, tracking, classification, and engagement. Field testing at Technology Readiness Experimentation (T-REX) 26-1 demonstrated 100% probability of defeat against Group 1 UAS targets with a mean engagement time of 6 seconds and 10 rounds per kill at ranges exceeding 160 meters. Operating in both autonomous and human-in-the-loop modes, Bullfrog achieved 99.45% operational availability while leveraging service-common M240B weapons and Modular Open Systems Architecture for rapid integration with Joint All-Domain Command and Control (JADC2) networks. At $300,000 per unit with $10 cost-per-engagement, Bullfrog demonstrates operational relevance, speed-to-field, and alignment with Army and Marine Corps autonomy priorities.
Cunningham, Jason, Clark, Alex
AI-Driven Predictive Sonification for Human-on-the-Loop Ground Vehicle Autonomy2026-01-75039/22/2026
Ground vehicle autonomy increasingly depends on human-on-the-loop (HOTL) supervision, yet supervisors are often overloaded by visual interfaces that can obscure emerging risks. This paper presents an AI-driven predictive sonification architecture that converts short-horizon forecasts of platoon behavior into structured auditory cues for supervisory monitoring. A forecasting engine predicts future vehicle interaction states and evaluates predicted and active violations to generate a composite risk indicator. When risk exceeds defined thresholds, a sonification module conveys risk magnitude and trajectory through changes in pitch, loudness, modulation, and spatial panning. The paper describes the system architecture, sonification design, operational use cases, and a planned human-subject evaluation. The proposed framework is intended to improve early awareness of emerging instability and support more timely supervisory intervention.
Plotzke, Zachary R., Mohammadi, Alireza, Cheung, Calvin M.
Development of a UAV Platform for Remote Inspection and Visualization2026-01-75119/22/2026
This paper details the development of an intelligence and inspection platform consisting of an attritable sub-250g UAV, a ground control station, and a visualization interface for users. The UAV architecture combines onboard obstacle detection and avoidance along with simultaneous localization and mapping to have full autonomous navigation inside of complicated GPS-denied environments. The ROS 2-to-Unreal Engine data pipeline allows for sensor fusion, data cleansing, and initial analysis as well as creation of a high-fidelity real-time 3D digital twin. The visualization interface allows users to easily identify critical features and turn data into intelligence to support decision making by soldiers and first responders.
Lee, Yeen K., Bainard, Sean, Shaughnessy, Michael, Bolger, Matt, Koepp, R. Tucker, Salehzadeh, Roya, Mallory, Stephen, Mynderse, James A., Guillen, Pedro, Hernandez, Margarita
View Translation: Geometry-Guided Latent Diffusion for Cross-View Image Synthesis on Ground Vehicles2026-01-75019/22/2026
Synthesizing novel camera views is important for autonomous ground vehicles, with applications in surround-view monitoring, occlusion recovery, and training data augmentation. We present View Translation, a geometry-guided latent diffusion framework that generates a target camera view from a source image, relative camera pose, and an available target-view depth prior. The method combines three components: a Vector Quantized Variational Autoencoder for compact latent encoding, a depth-based warping module that projects the source image into the target view to provide geometric guidance, and a ControlNet-augmented denoising UNet conditioned on source appearance, relative pose, and an auxiliary Image-Depth fusion network. Evaluated on KITTI and a simulated off-road dataset, our method achieves competitive FID while improving LPIPS and PSNR over baseline approaches, supporting cross-view synthesis for ground vehicle perception.
Mayekar, Omkar, Aiyetigbo, Mary, Salvi, Ameya, Samak, Tanmay, Samak, Chinmay, Desjardins, Brendan, Smereka, Jonathon, Brudnak, Mark, Krovi, Venkat, Luo, Feng, Li, Nianyi
The Impact of Observability Constraints on the Planning for Employment of Robotic and Autonomous Systems2026-01-75149/22/2026
Employment of Robotic and Autonomous Systems requires a different paradigm of mission planning. The GTRI Missioneer for Organic Collaborative Kill-Chains (MOCKS) effort was developed to mature the concepts of mission planning in light of autonomy, where the decision making is distributed across the battlespace and each individual entity has limited awareness of global state. The paper presents some initial characterizations of the impact of constraints on periods out of communication, or no-comms windows, on the number of resources required for missions of a certain operational area. This metric, along with the foundational metric of remaining effective range margin are foundational to robust a priori planning.
Spratley, Michael
Scene-Level Multi-View Instance Consistency Using Multi-Feature Vision Fusion for Synthetic Ground Vehicle Environments2026-01-75029/22/2026
Maintaining consistent object identities across multiple camera viewpoints is a critical challenge in synthetic perception environments used for autonomous ground vehicle evaluation. This paper presents a scene-level multi-view instance consistency framework that integrates OpenUSD scene composition, Omniverse Replicator synthetic-data generation, and a multi-feature vision fusion pipeline. The proposed approach combines semantic embeddings from CLIP, patch-level descriptors from DINOv2, geometric correspondences from LoFTR, mask-derived shape invariants using Hu moments, and relative-position priors to associate object instances across views, including visually identical objects. A compact composite scoring function fuses these complementary cues to achieve robust cross-view identity assignment while preserving OpenUSD asset modularity through grouped-prim support. Synthetic experiments across 120 multi-camera scenes demonstrate improved Top-1 Match Accuracy and Identity Consistency Rate, with reduced ID-switch occurrences compared to single-cue baselines. The framework supports scalable, repeatable, and traceable digital engineering workflows for defense-oriented perception evaluation.
Bhattacharya, Sambit, Nakamoto, Kyle
Intelligent Optimization Simulation Framework for Test and Evaluation of Autonomous Systems2026-01-75519/22/2026
This paper describes ongoing research and development of an efficient optimization/search–based modeling and simulation framework for rapidly identifying low-performance scenarios in advanced autonomous systems. Ensuring predictable, safe behavior across complex, integrated systems remains a core operational test-and-evaluation challenge. Our goal is to balance rigorous validation with timely deployment. We are developing TEAAS (Test & Evaluation of Advanced Autonomous Systems), a scalable, faster-than-real-time framework designed to uncover critical failure scenarios efficiently. Key features include GPU-accelerated parallel simulation and learning, computational intelligence–based search of optimal parameters, uncertainty quantification for reproducibility, and real-time physics-accurate sensor models. We conducted simulation experiments to evaluate and demonstrate the framework performance for two black-box ground-vehicle autonomous systems. Key results were that adequate uncertainty quantification can be achieved with as few as 10 repeated runs per simulation scenario, sensor realism has a significant effect on failure rate, distinct differences between the two autonomies failure modes were identified, and our efficient optimization/search methods identify critical performance regions in a small fraction of the number of simulations required by a naïve Monte Carlo search.
Snarski, S., Menozzi, A., Persons, B., Lazar, D., Khan, N.
Leveraging Standards-Based MOSA for SWaP-Constrained Deployable Systems with VNX+2026-01-75369/22/2026
®
Littlefield, Mark
SOSA Enabled Automotive Technologies Disrupt the Battlefield2026-01-75359/22/2026
.
Pilaud, William J.
Adopting Cots Technology for UGV Wheel Drive System2026-01-75619/22/2026
As the defense industry prioritizes speed of play to allow our warfighters to maintain a decisive edge over our adversaries, creativity is needed to leverage COTS effectively. This paper presents a case study of a fast-paced workflow leveraging modeling and simulation, targeted risk testing, thermal characterization, and accelerated life testing. Citation: K. May, J. Costa, J. Boyd, “Adopting COTS Technology for UGV Wheel Drive System,” In Proceedings of the Ground Vehicle Systems Engineering and Technology Symposium (GVSETS), NDIA, Novi, MI, Aug. 11-13, 2026.
May, Ken, Costa, Joao, Boyd, Jake
Stuck-Assist: A Large Multimodal Model-Based Stuck Recovery Assist for Driving on Off-Road Deformable Terrains2026-01-75139/22/2026
Recent advancements in off-road autonomy have shown significant progress in perception, planning, and control frameworks, including end-to-end learning approaches. Comprehensive results have been demonstrated in both simulation and real-world experiments; however, there are significant challenges in critical cases that need further evaluation. One such challenge is the immobilization of autonomous ground vehicles (AGVs) in unstructured off-road environments, which can significantly impact agriculture, space exploration, military operations, and search and rescue missions. Addressing this problem requires recovery strategies that are context-sensitive, adaptable to terrain and vehicle conditions, and effective in integrating multimodal inputs. To this end, this paper investigates the use of a large multimodal model (LMM) providing higher-level planning assistance with human-in-the-loop evaluations for vehicle recovery after immobilization in unstructured off-road terrain. The experimental simulation platform developed was based on the Algoryx (AGX) Dynamics engine for high-fidelity terramechanics interaction and vehicle physics combined with Unreal Engine 5. This platform was further integrated with a driving simulator equipped with steering wheel and pedal interfaces for human-in-the-loop experiments. We evaluated ten representative unstuck scenarios across two deformable terrains (loose sand and compact sand) under two modes: an unskilled baseline, where participants attempted recovery unaided, and a co-intelligence mode, where participants used LMM advisory instructions. The results show that LMM assistance improved stuck recovery rates by 70% compared to unaided and unskilled human driving.
Bhosale, Mayuresh, Whitson, Jordan A., Vahidi, Ardalan, Jia, Yunyi
Terrain-Aware Probabilistic Prediction of Suspension Component Forces in Unmanned Ground Vehicles2026-01-75099/22/2026
Unmanned ground vehicles (UGVs) operating in unstructured environments must account not only for terrain traversability but also for terrain-induced loads that affect component durability. Existing path planning approaches primarily consider obstacle avoidance and mobility, while neglecting cumulative structural degradation due to repeated loading. High-fidelity physics-based simulations can capture these effects but are computationally prohibitive for real-time applications. This study investigates machine learning-based surrogate models for predicting vehicle component reaction forces from terrain height sequences generated using a controlled parametric terrain formulation. Both feed-forward and recurrent neural network architectures are evaluated, and ensemble-based probabilistic techniques are incorporated to quantify predictive uncertainty. Results show that the ensemble long short-term memory (Ens-LSTM) model achieves the lowest prediction error (mean absolute error of 0.621 kN) while maintaining narrow 95% prediction intervals (3.22–4.28 kN). A simpler ensemble feed-forward network (Ens-NN) achieves comparable accuracy (0.626 kN) with reduced model complexity. These results demonstrate that data-driven surrogate models can provide accurate and uncertainty-aware force predictions, enabling the integration of structural reliability considerations into fatigue-aware path planning for UGVs.
Chua, Yang Kang, Mundiwala, Mohammad, Wang, Xudong, Castanier, Matthew, Hu, Zhen, Hu, Chao
Evaluating the Robustness of Segmentation Models against Adversarial Patches in Off-Road Environments2026-01-75209/22/2026
Semantic Segmentation (SS) is critical for autonomous vehicles to navigate off-road environments by identifying drivable terrain. Although models like ResNet34+UNet and EfficientViT have been proposed for these tasks, their susceptibility to localized adversarial patches in unstructured environments remains under-researched. This paper presents a comprehensive robustness evaluation of six real-time SS architectures, including the state-of-the-art YOLOv11 and YOLOv12 segmentation variants against five diverse adversarial patch schemes. Our experiments, conducted on a modified YCOR dataset, demonstrate that EfficientViT is the most resilient architecture, maintaining high accuracy with minimal performance degradation. In contrast, single-stage models like YOLOv11n-seg exhibit significant vulnerability, with pixel accuracy drops reaching 26.55%. We also show how decreases in overall segmentation accuracy impact the segmentation models’ ability to discern traversable terrain from non-traversable terrain.
Salas, Christopher, Pesé, Mert D., Li, Bing, Smereka, Jonathon, Cheng, Long
Real-Time Road Intelligence through a Hazard Authentication Model2026-24-00279/21/2026
Advanced Driver Assistance Systems (ADAS) are evolving beyond onboard perception. The ability to dynamically map and share temporary road hazards is important for connected and autonomous driving, but authenticating the data is critical. An in-vehicle hazard recognition layer performs real-time video analysis and geotagging on embedded platforms. Real-time video is analyzed for road hazards—such as potholes, construction zones, waterlogging, fallen trees, roadside accidents, and debris—using onboard cameras and lightweight computer vision models. This metadata is sent to a cloud-based aggregation layer to validate hazard reports. It is of paramount importance to validate hazard reports originating from diverse sources, regardless of their accuracy. This paper presents a mathematical approach to determine an overall Hazard Confidence Score (HCS) based on data received from diverse sources. A unique hazard authentication model is introduced to quantify the credibility of each hazard report using six validation metrics.
Bose, Souvik
Electro-Hydraulic Brake-by-Wire and Off-the-Shelf ABS Integration: Control Design and Hardware in the Loop Validation2026-01-07959/14/2026
This work presents the design of a control logic for an electro-hydraulic brake-by-wire in series with an off-the-shelf ABS unit for motorsport applications. Validation is performed through hardware-in-the-loop testing with a complete hydraulic layout, including the brake-by-wire actuator, the ABS module, and brake calipers. State of the art electro hydraulic brake-by-wire systems are increasingly adopted in top level motorsport and are now transitioning to high performance road vehicles, in combination with ABS and ESC. However, due to motorsport regulations, racing brake-by-wire systems do not incorporate ABS functionality. To combine the performance of motorsport grade actuators with the ease of use required for non professional drivers, a series configuration between brake-by-wire and ABS represents a natural solution. This architecture is also relevant for future road vehicle applications, offering additional redundancy for autonomous driving ready systems. A dedicated hardware-in-the-loop test rig has been developed to perform experimental testing of the complete brake system. Wheel dynamics are simulated in real-time using a single-axle vehicle model, and wheel speed signals are reproduced via a sensor emulator. Preliminary tests show that the original pressure-based brake-by-wire control strategy exhibits poor performance during ABS activation, as ABS operation significantly alters system behavior. To address this issue, an improved control strategy is proposed, introducing a dedicated control mode activated during ABS operation, with a smooth transition back to nominal control once ABS activity ceases. Experimental results demonstrate that the proposed strategy maintains closed-loop stability, avoids excessive pressure oscillations and piston end stop conditions, and, most important, does not interfere with ABS operation. Overall braking performance is fully preserved.
Milivinti, Massimiliano, Gimondi, Alex, Gobbi, Massimiliano, Cantoni, Carlo
Predicting Brake Squeal Occurrence Using Machine Learning with Temporal Feature Analysis of Braking-Torque Test Data2026-01-08129/14/2026
The automotive industry's paradigm shift toward autonomous driving and electrification has introduced new competitors threatening market dominance through differentiated value propositions. In this highly competitive landscape, delivering irreplaceable customer value requires providing sustainable and authentic luxury experiences. Quiet driving represents a tangible value that customers genuinely appreciate. Brake squeal—high-frequency noise arising from friction-induced vibration during braking—negatively impacts customer satisfaction and must be suppressed. Despite significant advances in brake squeal prediction modeling, the irregular nature of squeal generation mechanisms has prevented the development of a generalized predictive model applicable to product development processes. Development and verification remain largely experimental. This limitation constrains early-phase design validation, as brake squeal is highly sensitive to chassis and braking system design. When squeal issues emerge during post-design evaluation, fundamental improvements to pad materials become difficult. Consequently, damping characteristic tuning is employed for mitigation, incurring substantial development costs. This study addresses this challenge through systematic feature engineering of time-series braking data—brake torque, disc rotational speed, disc temperature, and brake pressure—collected during squeal evaluation tests. Based on the hypothesis that environmental conditions and brake system characteristics influence mechanical behavior, time-series features exhibiting strong predictive association with squeal occurrence were derived, and a machine learning model was developed to predict squeal occurrence probability using these features as input variables. The model's predictive performance was validated by comparing squeal probability predictions derived from independent torque performance evaluation data against actual squeal evaluation results. This validation confirms that the model successfully predicts squeal occurrence probability from dynamometer torque performance data alone. Consequently, this approach enables the prediction of squeal occurrence probability in early development phases before formal noise assessment is conducted, streamlining the development process and significantly reducing verification costs while contributing to quieter driving experiences.
Cho, Sunghyun, Yoon, Jungro, Kim, Yoon Cheol, Kim, Jeongkyu, Kim, Sungho, Baek, SongYi, Kim, Won Joon, Choi, Kyung Rok
Optimization of Order Picking in an AMR-Assisted Picker-to-Parts System2026-99-15679/11/2026
With global retail sales expanding and same-day delivery demand on the rise, efficient order picking operations in warehouses have become critical to success. To improve order picking processes, warehouse managers increasingly rely on autonomous mobile robots (AMRs), which improve the performance of traditional picker-to-parts systems. This paper investigates an AMR-assisted picker-to-parts system in which a set of customer orders must be fulfilled. The orders are first batched, and the resulting batches are assigned to individual pickers. Each picker works in a batch-by-batch manner, manually retrieving items from picking aisles and handing over the completed batch to an AMR waiting at the cross aisle. After receiving a full batch, the AMR transports it to the designated depot before returning to serve the next batch. The objective is the minimization of the total tardiness of all orders. The problem is formulated as a mixed-integer programming (MIP) model, and several effective heuristic algorithms are developed. Extensive computational experiments are conducted to evaluate the performance of the proposed algorithms and compare them with a commercial MIP solver.
Jin, Bo, Peng, Jianxin
Adaptive MPC Path Tracking Control for Autonomous Vehicles2026-99-15709/11/2026
This work introduces a Model Predictive Control (MPC) path tracking controller, which is developed to boost robustness, tracking precision, and vehicle stability when navigating high-speed and high-curvature driving scenarios. First, a 3-degree-of-freedom (3-DOF) dynamic model of the vehicle is established to serve as a reference. Next, a linear time-varying MPC control algorithm is formulated, with constraints on yaw rate, lateral velocity, and road boundary conditions taken into account; a comprehensive performance metric that balances tracking accuracy and control smoothness is also defined. Third, the time-domain parameters of the MPC framework are optimized using an improved genetic algorithm. Finally, the effectiveness and accuracy of the proposed controller are validated via co-simulation experiments conducted on the Matlab/Simulink and Carsim platforms. Simulation results demonstrate that the controller exhibits excellent robustness: the peak lateral tracking error is only 0.05 m on high-friction roads and 0.12 m on low-friction roads, with a maximum heading error of 0.15°. Additionally, the vehicle’s dynamic stability is notably enhanced: the yaw rate is reduced by 9.6% and 15.7% on high- and low-adhesion roads, respectively, while the sideslip angle is decreased by 13.2% and 18.4% under the same conditions.
Yu, Hanzhengnan, Hou, Xiaoyi, Zhang, Hao, Zhou, Weichen, Liu, Yu
Affine Formation Control of Multi-Agent Systems with Switching Topologies2026-99-11289/11/2026
Tackling the heavy computation of affine formation control under switching topologies—rooted in frequent stress matrix recalculation—this paper presents a distributed control framework fusing consistency estimation with dynamic error constraints for efficient coordination. In a leader-follower architecture, affine transformation parameters are estimated by followers using local information—global stress matrix solutions are thus avoided. A time-varying constraint function and Lyapunov stability analysis are devised to ensure tracking errors converge to specified accuracy within a predetermined time. Both theoretical analysis and simulation results show that this method greatly simplifies computation. It also supports flexible formation transformations such as translation and scaling, making it a stable and reliable solution for dynamic scenes.
Liu, Guicai, Li, Jianzhen, Zhou, Junyi, Tang, Jiye
A Spatio-Temporal Graph Network with Informer-TCN Fusion for Vehicle Trajectory Prediction on Highways2026-99-11049/11/2026
Accurate vehicle trajectory prediction is essential for the driving safety and efficiency of autonomous vehicles. However, this task remains challenging due to the complex spatial interactions among traffic participants and the wide range of temporal dependencies in motion sequences. To address these issues, this paper proposes a novel hybrid deep learning framework suitable for cloud-based control platforms, providing a foundational algorithmic solution for vehicle-infrastructure cooperative perception and decision-making. The proposed architecture employs an Adaptive Graph Convolutional Network (AGCN) to adaptively learn spatial relationships and interactions among vehicles. It also utilizes the Informer model, known for its efficient ProbSparse self-attention mechanism, to capture long-term temporal dependencies in trajectory sequences. Furthermore, a Temporal Convolutional Network (TCN) is integrated to enhance the model’s ability to learn fine-grained local temporal features. The proposed model is evaluated on the NGSIM dataset. The dataset is chronologically ordered and split into training (80%), validation (10%), and testing (10%) sets. Experimental results show that the proposed method achieves an average minADE of 2.032 m and minFDE of 2.866 m over a 5-second prediction horizon, outperforming several baseline models such as LSTM, CNN-LSTM, and Social-GAN. These results indicate the effectiveness of the AGCN-Informer-TCN combination for trajectory prediction. The study suggests potential for integration into intelligent transportation cloud control platforms.
Liang, Ziyan, Yuan, Rui, Zhou, Pengying, Yang, Shu, Li, Weidong, Zhang, Zijian
Research on an Evaluation Method and Application of Autonomous Vehicles in Traffic Flow Environment2026-99-15629/11/2026
Traffic flow environment testing is indispensable in the research and evaluation process of intelligent vehicles. However, the current vehicle evaluation systems mostly focus on simple dynamic scenarios and lack a comprehensive assessment of vehicle performance in complex traffic flow environments. In view of this, this paper constructs a multi-dimensional comprehensive performance evaluation system for vehicles in traffic flow environments. Firstly, based on the randomness and dynamics of traffic flow, four core evaluation dimensions covering safety, comfort, efficiency, and economy are constructed. In terms of security, the collision time and Predicted Safety Metrics are adopted. This enables the dual quantification of immediate collision risks and dynamic obstacle avoidance capabilities. The comfort evaluation focuses on the impact of vibration frequency on the human body and constructs a graded quantitative index. Efficiency and economy evaluation models are built through time cost and energy consumption cost. The verification is carried out under different traffic flow scenarios, and the final results demonstrate the consistency between the evaluation method of this paper and the expert evaluation method, thereby verifying the rationality of the evaluation system presented in this paper.
Wang, Guangyu, Song, Shiping, Zhang, Cheng, Qu, Ge, Yu, Xiaojun
Research on High-Reliability Testing Systems for Intelligent and Connected Vehicle2026-99-11119/11/2026
With the large-scale application of intelligent connected vehicles, the verification of their functional safety and reliability has become a core bottleneck in the industrial development. The traditional real- vehicle road test method can no longer meet the current demand for large-scale test verification due to problems such as high cost, low efficiency, and difficulty in reproducing dangerous scenarios. This paper studies the vehicle-in-the-loop simulation test system based on a digital twin. By constructing a virtual scenario highly consistent with the real world, physical-level multi-source perception signals are simulated and mapped to the system under test to enable high- reliability verification of real vehicles. In terms of lateral and longitudinal control functions, multiple sets of test cases are selected respectively for comparison between road tests and virtual simulation tests. The results show that the accuracy of key indicators is above 90%, which provides practical reference for the subsequent test and verification system of high-level autonomous driving.
Hou, Quanshan, Tang, Ke, Gao, Tian, Chen, Tao, Zhou, Si
Emotional Impacts of Manual and Assisted Driving Modes under Varying Time Pressures: A Scenario-based Simulation Study2026-99-15089/11/2026
Emotion as a critical psychological variable in driving behavior has been extensively confirmed to exert a profound influence on traffic safety. With the advancement of autonomous driving technologies, the application value of emotion regulation mechanisms in intelligent transportation systems is gaining importance, given their proven impact on reducing unsafe driving tendencies and improving human-machine interaction quality. However, systematic research on the mechanisms underlying emotional variation across different driving modes remains relatively scarce. This study developed a questionnaire based on the Chinese version of the Driving Anger Scale (DAS), incorporating 19 traffic scenarios and corresponding video stimuli, to examine the impact of manual and assisted driving on driving anger under varying levels of time pressure. In addition, the moderating effects of individual factors such as age and education were analyzed. Descriptive statistics, independent samples t-tests, and one-way ANOVA were applied to 113 valid responses to construct a three-dimensional analytical framework involving driving mode, time pressure, and demographic variables. The results indicate that under high time pressure, assisted driving systems significantly reduce drivers' anger levels in typical delay-related scenarios compared with manual driving. Age is identified as a key moderating factor influencing emotional responses. This research provides theoretical support for the development of emotion-aware driving intervention systems and offers empirical evidence for the formulation of future strategies for driving emotion regulation.
Zhang, Tingjia, Li, Ruiheng, Zhu, Tong
A Vision-Based Vehicle Collision Warning Algorithm Integrating YOLOv8 and DeepSORT2026-99-11079/11/2026
With the continuous development of autonomous driving technology, vehicle collision warning systems are playing an increasingly important role in this field, promoting the progress and improvement of the whole autonomous driving field. But existing methods have low detection rates and unstable multi-target tracking performance. In particular, the estimation of relative motion parameters is still inaccurate due to the loss of direction information when relative speed is described as a scalar quantity. These limits produce easy collisions in the judgment of the car in a dangerous traffic environment. To solve these problems, a vehicle collision warning algorithm based on YOLOv8 and DeepSORT is proposed in this paper. YOLOv8 is introduced to detect the vehicle precisely, and DeepSORT is used to enhance multi-target vehicle tracking. The geometric principles of monocular vision are applied to extract key motion parameters such as distance and direction-signed relative speed. A classification logic is designed to distinguish between positive and negative relative velocities, enabling more accurate judgment of collision risk levels. In order to further enhance the reliability of the system, a four-stage cascaded false-alarm suppression mechanism is proposed. By adding velocity direction validation, distance validity checks, adaptive confidence thresholds, and a temporal consistency verification mechanism, the false alarm rate is reduced, and the proposed approach can realize direction- aware velocity estimation without requiring additional sensors and can be easily integrated into the existing YOLOv8 perception system.
Qin, Xiaoyu, Li, Wei
3D Dynamic Object Detection with Multi-Modal Temporal Fusion2026-99-11179/11/2026
This research aims to address the critical challenge of accurately detecting and estimating the state of dynamic objects in autonomous driving. Traditional 3D object detection methods often struggle with motion perception, particularly in velocity estimation, due to the lack of information in single frame perception. We propose a novel framework that enhances the BEV representation with temporal modeling. The core of our method is a two-stage temporal fusion process. First, we align historical BEV features to the current coordinate frame to eliminate the interference of ego-motion. Subsequently, a dedicated temporal fusion encoder, architected with residual connections and a Feature Pyramid Network, refines the aligned multi-frame BEV features to capture complex motion patterns and improve multi-scale object representation. This approach directly tackles the problem of motion decoupling. By aligning features, we disentangle object motion from ego-motion. The temporal fusion encoder then mitigates the positional ambiguity of moving objects in the fused BEV space, a common issue in simple feature concatenation, leading to more robust detection. We built a dataset following the structure of the nuScenes dataset, using data collected from an autonomous driving simulation platform. The evaluation results on our simulation dataset demonstrate that the proposed temporal module achieves a 13.0% improvement in NDS score and a substantial 29.7% reduction in velocity error (mAVE). These results demonstrate that our temporal fusion strategy effectively enhances 3D detection accuracy in dynamic scenarios.
Shao, Mengjia, Li, Wei, Bai, Jie, Zhu, Shaoxiong, Xu, Chenjie
Analysis of Latency-Influencing Factors in Vehicle-Road-Cloud Integrated Systems Based on Fault Tree Analysis and Analytic Hierarchy Process2026-99-15439/11/2026
Vehicle–road–cloud integrated systems have great potential in terms of improving traffic efficiency and achieving intelligent automatic driving through the integration of on–board terminals, roadside facilities, and cloud computing. However, their operational capabilities are heavily reliant on ultra-low-latency collaborative communication. This paper constructs a latency fault tree model to comprehensively analyze multi-source triggering paths of computation delay and reveals the formation mechanism of the delay path from “germination–induction–evolution”. On this basis, the Analytic Hierarchy Process is used to construct a three-level evaluation framework, and the influencing factors are quantitatively evaluated using NS-3 simulation data of the 004-V2X Communication Performance Testing Dataset. The result shows that the weight value of the network communication layer is the largest, 0.498, which shows that the bottleneck of performance in network communication is the wireless link quality. The cloud processing layer is second 0.327, which is dominated by the computational complexity and resource allocation policy. The impact of the onboard terminal layer is the smallest, 0.175. The FTA–AHP framework supported by empirical data can find the key factors affecting delay, which can help engineering optimization. It is noted that the AHP consistency check (CR) just checks the inner transitivity of expert judgment (i.e., the matrix consistency), while it cannot assure the objectivity and the bias elimination. We reduce the subjectivity by combining multiple experts, anchoring judgment with the simulation data, and performing a sensitivity check on the perturbation of the weights.
Xu, Yunchuan, Wang, Xiaomeng, Wang, Yan
SAE TOMORROW TODAY - Liability in the Age of Autonomous Vehicles135839/8/2026
When AI is driving, who's responsible? Listen in as we sit down with Jennifer Dukarski, Emerging Tech, Media, and Intellectual Property Shareholder at Butzel, to unpack the rapidly evolving legal landscape surrounding SDVs and AVs. From AI decision-making and over-the-air updates to data retention, cybersecurity, consumer education, and the growing complexity of liability, we explore how existing legal frameworks are being tested by next-gen mobility. Whether you're developing autonomous technology, managing risk, or shaping industry standards, this conversation offers an insightful look at the legal challenges that will define the future of transportation. Have your own thoughts on this topic? We'd love to hear from you! Share your comments, questions and ideas for future topics and guests to podcast@sae.org. Don't forget to take a moment to follow SAE Tomorrow Today--a podcast where we discuss emerging technology and trends in mobility with the leaders, innovators and strategists making it all happen--and give us a review on your preferred podcasting platform. Follow SAE on LinkedIn, Instagram, Facebook, X, and YouTube. Follow host Grayson Brulte on LinkedIn, X, and Instagram.
Patterson, Lori
EDITORIAL: Where to? The challenges of automated driving around the world26AUTP09_059/1/2026
Engineer them to identical and tough specs all you want; autonomous vehicles are unlikely to drive the same way everywhere. They may all have the same technical capabilities, but this world is too messy for an AV engineer in San Francisco to be working on the same problems as one in Amsterdam. These global differences were front and center at the recebt 2026 SAE Automated Transportation Symposium (see more on page 11). I came away with renewed excitement about automated driving technology, given the real-world impact AVs are already making. Reports from Waymo, Kodiak and Uber and others all the way down prove the technology is working. But uccessful deployment depends on more than vehicle technology.
Blanco, Sebastian
A New Research Facility for Robotics and Alloy Discovery in Texas26AERP09_099/1/2026
Researchers have received a $24.9 million grant to develop an NSF Materials Innovation Platform where universities, national laboratories and industry members can run alloy-discovery campaigns using robotic systems and artificial intelligence. Texas A&M University, College Station, Texas A six-year, $24.9 million grant from the U.S. National Science Foundation (NSF) will establish the Autonomous Robotic Metallurgist Materials Innovation Platform (ARM-MIP) at Texas A&M University - a national user facility where robots and artificial intelligence take on the repetitive labor of alloy discovery, freeing scientists to focus on discovery itself. ARM-MIP will be sited at The Texas A&M University System RELLIS Campus in Bryan, a dedicated research campus that also hosts the Texas A&M Engineering Experiment Station and the U.S. Army Transformation and Training Command's central testing hub at the George H.W. Bush Combat Development Complex. The platform is designed to reach more than 200 users a year and 50 alloys per month in its first year, scaling to more than 200 alloys per month by year six.
Resilient Trajectory Tracking Static Output-Feedback Control for Automated Vehicles under Denial-of-Service Attacks10-11-01-00028/24/2026
The safety and reliability of autonomous vehicles are critically linked to network communication quality. Consequently, threats such as Denial-of-Service attacks pose significant risks by disrupting communication between essential components and jeopardizing driving safety. This article presents a robust trajectory tracking control method resilient to such attacks. Utilizing a Linear Parameter–Varying control framework, the method effectively addresses time-varying vehicle speeds and incorporates a norm-bounded strategy to manage tire behavior uncertainties. By employing a Static Output-Feedback scheme, it avoids the need of costly sensors while maintaining a straightforward control structure suitable for real-time implementation. Using Lyapunov’s method, we analyze the system stability under attack conditions, ensuring the controller remains robust against disturbances. The design of the resilient controller is formulated as a convex optimization problem with Linear Matrix Inequalities constraints. The effectiveness of the proposed controller is validated through co-simulation in MATLAB/CarSim®, where it outperforms several state-of-the-art controllers across different driving scenarios, maintaining consistent tracking performance despite varying attack severity levels.
Meléndez-Useros, Miguel, Viadero-Monasterio, Fernando, Nguyen, Anh-Tu, López-Boada, María Jesús
Items per page:
1 – 50 of 3365