Browse Topic: Lidar

Items (444)
Find
Landing Site Selection and Visual Scene Generation Based on LiDAR Point Cloud2026-99-18197/17/2026
This paper focuses on autonomous drone landing scenarios. Addressing the core requirements of accurate landing site assessment and intuitive visual presentation, it conducts in-depth research on the application of 3D LiDAR (TOF technology) point cloud data. LiDAR captures point cloud data containing 3D coordinates and reflection intensity values. While sparse, non-uniform, and disordered, its high measurement accuracy and strong anti-interference capabilities make it a key sensor for landing terrain perception. Based on a review of recent research results from related teams, this study designed and implemented a comprehensive technical solution: First, raw point cloud data is acquired via the UDP protocol combined with an SDK interface. Preprocessing is then performed using voxel grid filtering (downsampling) and radius filtering (denoising). The assessment area is then divided into a row-by-column grid. A sliding window method is used to calculate the elevation difference, empty grid ratio, flatness, and slope of each grid. Based on these attributes, the grids are classified into six categories: Risk, Warning, Blank, Unknown, No Landing, and Landing. Finally, a grid attribute coloring method and OpenGL 3D rendering are used to generate the visual scene. Through the development of verification programs and moving obstacle experiments, it has been proven that the solution can efficiently process point cloud data and accurately identify safe landing areas, providing key technical support for the engineering realization of the autonomous landing function of drones, and also laying the foundation for the intelligent development of drone landing decisions in complex environments.
Guo, Hangyu, Shi, Zhe
Validation and Exploration of Multimodal Deep-Learning Camera–LiDAR Calibration Models2026-01-50377/13/2026
This paper presents an innovative study in exploring, evaluating, and implementing deep-learning architectures for the calibration of multimodal sensor systems. The aim of this paper is to leverage the use of sensor fusion to achieve dynamic, real-time alignment between 3D LiDAR and 2D camera sensors. Static calibration methods are tedious and time-consuming, which is why we propose utilizing conventional neural networks (CNNs) coupled with geometrically informed learning to solve this issue. We leverage the foundational principles of extrinsic LiDAR–camera calibration tools such as RegNet, CalibNet, and LCCNet by exploring open-source models that are available online and compare our results with their corresponding research papers. Requirements for extracting these visual and measurable outputs involved tweaking source code, fine-tuning, training, validation, and testing of each of these frameworks for equal comparisons. This approach aims to investigate which of these advanced networks produces the most accurate and consistent predictions. Through a series of experiments, we reveal some of their shortcomings and areas for potential improvements. We find that LCCNet yields the best results among all the models that we validated.
Karramreddy, Venkat Sai Raxit, Mitchell, Liam
Microvision fast-tracks adoption of Luminar tech; eyes defense, industrial sectors26TOFHP06_096/1/2026
In the two months since Microvision bought Luminar and acquired key tech and talent, the sensor company has been busy. In that time, they've merged key lidar units from each company and created a perception software stack to run it in a convincing demo of its ADAS and autonomous capabilities. The company is also pushing innovative lidar tech into the defense drone and antidrone markets, already working with a German defense supplier that works with NATO member countries.
Clonts, Chris
Diffusion Framework with Cross-Modality Fusion Perception for Autonomous Driving in Urban Traffic2026-99-07495/15/2026
End-to-end autonomous driving in urban environments faces three core challenges. First, camera and LiDAR sensor heterogeneity causes cross-modal perception inconsistencies and sensor fusion instability. Second, diffusion models suffer from training instability due to scale variance and distribution changes, which limits generalization. Third, traditional trajectory decoders lack structured interaction with semantic elements, thereby undermining planning rationality. To address these issues, CMFPNet introduces an integrated framework with three key modules. The HGCF-Backbone integrates LiDAR and camera features using channel focus, deformable cross-focus, and state space modeling to enhance semantic alignment. The NST module maps physical trajectories to normalized space, employing truncated diffusion sampling for stable generation in just 2–4 steps. The NDA models trajectory generation as a semantic narrative, utilizing a six-stage semantic attention flow incorporating BEV context, interactive dynamics, and self-states. Experiments on the NAVSIM dataset demonstrate CMFP Net’s superiority over existing baselines, showing outstanding generalization and trajectory stability in challenging scenarios. Notably, the truncated sampling strategy achieves an 8–10× acceleration during inference while maintaining decision accuracy and reducing computational costs. CMFPNet provides a scalable, semantically consistent solution for diffusion-based autonomous driving with significant potential in both research and practical deployment.
Qu, Yanwei, Mo, Hangjie
Flight Test Results for Velocity Aided Navigation in GPS Denied EnvironmentsF-0082-2026-01295/5/2026
This paper presents the flight-test evaluation of a velocity-aided navigation solution that integrates inertial measurements with line-of-sight (LOS) Doppler velocity observations from the Psionic Navigation Doppler Lidar (PNDL) prototype to support navigation in GPS-denied environments. LOS velocity measurements collected during a helicopter flight-test campaign were first compared with velocities derived from an Applanix reference navigation system to assess measurement accuracy. The navigation solution was then developed and evaluated under simulated GPS-denied conditions by removing GPS aiding and continuing operation using LOS velocity measurements alone for extended periods. Results show that Doppler lidar velocity aiding effectively constrains inertial navigation error growth and maintains a stable navigation solution during prolonged GPS outages. These flight-test results demonstrate the utility of FMCW Doppler lidar velocity measurements as an enabling technology for Assured Positioning and Navigation (APN) and underscore its applicability to Contested Logistics operations, where resilient, GPS-independent navigation is essential for mission continuity.
Hull, Jason, Pierrottet, Diego, Monaco, Jeffrey
A Comparison between Leica RTC360 and Recon-3D Scan Data in Pre- and Post-Collision Vehicle Documentation2026-01-50304/22/2026
Documenting and mapping using three-dimensional (3D) technologies have become essential in crime- and crash-scene investigations in recent years. Traditionally, this has been accomplished using terrestrial laser scanners (TLS), which often come with significant upfront costs. In contrast, Recon-3D, launched in 2022, leverages the capabilities of Apple’s light detection and ranging (LiDAR) sensor, available in Pro and Pro Max models since 2020. This study aims to evaluate the relative accuracy of documenting vehicles in both pre- and post-collision conditions using these technologies. A deviation analysis was conducted utilizing CloudCompare software to compare point cloud data collected from the Leica RTC360 laser scanner with that obtained from Recon-3D for 7 vehicles in a pre- and post-impact condition for a total of n = 14 vehicles. At the 1, 2, and 3 cm deviation thresholds, the average percent of points which fell below each threshold level for all vehicles was 66%, 91%, and 97%, respectively. Overall, the results indicate that Recon-3D delivers point cloud data that is useful for pre- and post-collision vehicle documentation.
Lim, Jihwa, Liscio, Eugene
Enhancing LiDAR Performance under Fog with Adaptive Multi-Echo and Feature-Based Filtering2026-01-00114/7/2026
LiDAR (Light Detection and Ranging) systems are essential for autonomous driving (AD) and advanced driver-assistance systems (ADAS), providing accurate 3D perception of the surrounding environment. However, their performance significantly deteriorates under adverse weather conditions such as fog, where laser pulses are scattered by airborne particles, resulting in substantial noise and reduced ranging accuracy. This scattering effect makes it difficult to detect objects within or behind particulate matter, posing a serious challenge for reliable perception in real-world driving scenarios. To address this issue, we propose an algorithm that combines adaptive multi-echo signal processing with a feature-integrated, rule-based denoising framework to enhance LiDAR performance in noisy environments. The multi-echo approach selectively utilizes meaningful signal returns by evaluating both intensity and relative echo positions. Based on predefined rules, the algorithm identifies the echo most likely to represent a real object. The rule-based denoising algorithm dynamically adjusts thresholds by integrating multiple features, including point clouds density, intensity, and echo width. These features are evaluated in conjunction with measured distance to adaptively suppress fog noise and improve signal reliability. This synergistic method enables robust detection of real objects even in low-visibility conditions. Experimental evaluations demonstrate that the proposed algorithm significantly improves effective ranging distance under adverse conditions compared to conventional methods. Furthermore, it eliminates up to approximately 99% of noise induced by airborne particles in foggy scenarios. These results highlight the potential of our approach to enhance LiDAR reliability and safety in real-world automotive applications, contributing to the advancement of autonomous driving technologies under all-weather conditions.
Kaito, Seiya, Zheng, Shengchao, Fujioka, Ibuki, Beppu, Taro
A Comparison of Road Grade Preview Signals from Lidar and Maps2026-01-00264/7/2026
Road grade can impact the energy efficiency, safety, and comfort associated with automated vehicle control systems. Currently, control systems that attempt to compensate for road grade are designed with one of two assumptions. Either the grade is only known once the vehicle is driving over the road segment through proprioception, or complete knowledge of the oncoming road grade is known from a pre-made map. Both assumptions limit the performance of a control system, as not having a preview signal prevents proactive grade compensation, whereas relying only on map data potentially subjects the control system to missing or outdated information. These limits can be avoided by measuring the oncoming grade in real-time using on-board lidar sensors. In this work, we use point returns accumulated during travel to estimate the grade at each waypoint along a path. The estimated grade is defined as the difference in height between the front and rear wheelbase at a given waypoint. Kalman filtering techniques are used to mitigate the effects of odometry and motion uncertainty on the grade estimates. This estimator’s performance is compared to the measurements of a map created with a GNSS/INS system via a field experiment. When compared to the map-based system, the lidar-based estimator produces an unbiased error with a standard deviation of 0.6 degrees at an average range of 52.7 meters. By having similar precision to map-based systems, automotive lidar-based grade estimation systems are shown to be a valid approach for measuring road grade when a map is unavailable or inaccurate. In using lidar as an input signal for grade-based control system tasks, autonomous vehicles achieve higher redundancy and independence in contrast to existing methods.
Schexnaydre, Logan, Poovalappil, Aman, Robinette, Darrell, Bos, Jeremy
Self-Supervised Gaussian-Based Framework for Robust Static-Dynamic Decomposition and Real-Time Urban Scene Reconstruction2026-01-00104/7/2026
High-fidelity 3D reconstruction of large-scale urban scenes is critical for autonomous driving perception and simulation. Existing neural rendering methods, including NeRF and Gaussian-based variants, often face challenges like unstable geometry, noisy motion segmentation, and poor performance under sparse viewpoints or varying illumination. This paper presents a self-supervised Gaussian-based framework to address these challenges, enabling robust static–dynamic decomposition and real-time scene reconstruction. The proposed method introduces three innovations: (1) a semantic–geometric feature fusion module that combines semantic context and geometric cues for reliable motion prior estimation; (2) a cross-sequence geometric consistency constraint that enforces depth and surface continuity across time and viewpoints; (3) an efficient Gaussian parameter optimization strategy that stabilizes geometry by jointly constraining scale and normal updates. Experiments on the Waymo Open Dataset and KITTI benchmarks show that the proposed framework improves PSNR by up to +1.3 dB, SSIM by +0.015, and reduces depth L1 error by over 25%, while achieving real-time rendering speeds exceeding 40 FPS. These results demonstrate that the proposed framework provides a robust, scalable solution for urban scene reconstruction, with practical applications in autonomous driving.
Feng, Runlei, Wang, Ning, Zhang, Zhihao
Development of Technology to Distinguish Contaminants using LiDAR Near-Field Point Cloud based on Real-Field Data2026-01-00204/7/2026
Reliable environmental perception under adverse and contaminated conditions is a critical requirement for autonomous driving systems. Although LiDAR sensors play a central role in such perception, their performance is significantly degraded by surface contamination caused by environmental factors such as rain, snow, dust, anti-icing materials, and bug splatter impacts. However, most existing public datasets and prior studies rely on simulated or laboratory-generated contamination scenarios, which limit their applicability to real-world autonomous driving. To address this gap, we construct a large-scale real-world dataset collected from approximately 22,000 km of on-road driving across diverse regions of the United States, covering a wide range of naturally occurring environmental contamination conditions. The dataset was acquired using a multimodal sensing platform integrating LiDAR, perception RGB cameras, infrared camera sensors, and external monitoring systems, enabling comprehensive observation of sensor behavior under realistic operating environments. Based on this dataset, we propose a scalable contaminant classification framework that focuses on LiDAR surface contamination. A key contribution of this study is the introduction and exploitation of near-field point cloud features, which capture backscattered laser signals caused by surface contamination and exhibit a strong correlation with contamination severity and type. Using raw LiDAR signals, we utilize sixteen feature functions and train supervised learning models to classify seven distinct contaminant categories. Experimental results demonstrate that the proposed approach achieves classification accuracy exceeding 95% under real-world driving conditions, significantly outperforming prior laboratory-based studies. Furthermore, the framework is designed for practical deployment and can be extended to additional contaminant types and geographic regions through incremental data collection and learning. The proposed methodology enables real-time identification of LiDAR contamination sources, providing a critical foundation for adaptive sensor-cleaning strategies. By supporting contamination-aware sensor maintenance, this work contributes to cost- and weight-efficient sensor system design and represents an essential step toward achieving reliable Level 4 autonomous driving.
Kim, Hunjae
Hierarchical 3D LiDAR Localization Improves Robot PositioningTBMG-548884/1/2026
Mobile robots must continuously estimate their position to navigate autonomously. However, satellite-based navigation systems are not always reliable: signals may degrade near buildings or become unavailable indoors. To operate safely and efficiently, robots must interpret their surroundings using onboard sensors and robust localization algorithms.
Vehicle Self-Soiling from Exterior Cleaning Mechanisms2026-26-04251/16/2026
The automotive industry is rapidly advancing towards autonomous vehicles, making sensors such as Cameras, LiDAR, and RADAR critical components for ensuring constant information exchange between the vehicle and its surrounding environment. However, these sensors are vulnerable to harsh environmental conditions like rain, dirt, snow, and bird droppings, which can impair their functionality and disrupt accurate vehicle maneuvers. To ensure all sensors operate effectively, dedicated cleaning is implemented, particularly for Level 3 and higher autonomous vehicles. It is important to test sensor cleaning mechanisms across different weather conditions and vehicle operating scenarios to ensure reliability and performance. One crucial aspect of testing is tracking the trajectory of the cleaning fluid to ensure it does not cause self-soiling of vehicles and affects the field of view or visibility zones of other components like the windshield. While wind tunnel tests are valuable, digitalizing this process is vital for making design decisions early in vehicle development. This work presents a digital methodology to test the self-soiling of a vehicle due to the cleaning systems present on vehicle exterior components, e.g. during mud cleaning at different vehicle speeds. The cleaning mechanism involves multiple water nozzles positioned above, below, or on the sides of these components, which spray water jets to remove dirt or mud deposits. The developed numerical method models the motion of cleaning fluid and contaminants after component cleaning. Steady-state aerodynamic simulations using the Finite Volume Method (FVM) are used to capture airflow, while the interaction of air with cleaning fluid and components is analyzed using a Smoothed Particle Hydrodynamics (SPH) solver. Correlations from this study and wind tunnel tests reveal potential optimization opportunities for existing cleaning systems by inspecting surrounding airflows at various vehicle speeds. Preliminary design evaluations indicate a specific vehicle speed range where self-soiling of vehicle components such as the windshield occurs due to mud cleaning. The proposed numerical method provides the capability to evaluate and qualitatively compare vehicle self-soiling due to various cleaning system designs of exterior components, offering valuable insights for optimizing cleaning mechanisms in autonomous vehicles.
Mane, Suvidya, Makam, Sri Lalith Madhav, Varghese, Rixson, Desu, Harsha
Realistic LiDAR Data Simulation for Autonomous Systems using Physics-Informed Learning2026-26-01381/16/2026
Accurate and realistic simulation of LiDAR data is critical for the development and validation of autonomous driving systems. However, existing simulation approaches often suffer from a significant sim-to-real gap due to oversimplified modelling of physical interactions and environmental factors. In this work, we present a physics-informed deep learning framework that bridges this gap by enhancing the realism of simulated LiDAR data using generative adversarial networks guided by domain-specific physical constraints for LiDAR intensity. Our method incorporates key physical factors such as range, surface material properties, angle of incidence, and environmental conditions along with their underlying physical relationships as constraints into the Cycle-Consistent GAN architecture, enabling it to learn realistic transformations from synthetic to real-world LiDAR intensity data without requiring paired samples. We demonstrate the effectiveness of our approach across multiple datasets, showing consistent improvements in statistical similarity metrics and downstream perception tasks such as semantic segmentation. The proposed algorithm has been integrated into the Sim-DaaS simulation engine, providing a robust tool for the research and industrial community to generate high-fidelity LiDAR data for training and evaluation of autonomous systems.
Anand, Vivek, Yadav, Sourav, Limba, Mohit, Pandey, Gaurav, Lohani, Bharat
LLS-SMGSC: A Lidar Localization System Based on Simplified Maps Integrated with Global Search Capabilities2025-01-730312/31/2025
To address the issues of large storage requirements in maps and the dependence of localization accuracy on initial pose estimation, this paper proposes a novel relocalization method named LLS-SMGSC, which is based on simplified maps integrated with Global Search capabilities. Firstly, we partition the map-based on grid size to reduce memory usage. Next, we voxelize the point cloud and map and extract surfel. Then, a coarse-to-fine hierarchical alignment module between the initial frame and maps to estimate the initial global pose. Finally, unmanned platform pose is estimated by the Normal Distribution Transform (ndt) algorithm. Experiments demonstrate that LLS-SMGSC achieves the highest localization accuracy in both unstructured and structured environments while maintaining computational efficiency.
Quan, Zhiheng
3D LiDAR for Improved Rover Traversal and ImageryTBMG-5429512/1/2025
Planetary and lunar rover exploration missions can encounter environments that do not allow for navigation by typical, stereo camera-based systems. Stereo cameras meet difficulties in areas with low ambient light (even when lit by floodlights), direct sunlight, or washed-out environments. Improved sensors are required for safe and successful rover mobility in harsh conditions. NASA Goddard Space Flight Center has developed a Space Qualified Rover LiDAR (SQRLi) system that will improve rover sensing capabilities in a small, lightweight package. The new SQRLi package is developed to survive the hazardous space environment and provide valuable image data during planetary and lunar rover exploration.
VORA™: Vision for Off-Road Autonomy2025-01-04419/16/2025
The Vision for Off-road Autonomy (VORA) project used passive, vision-only sensors to generate a dense, robust world model for use in off-road navigation. The research resulted in vision-based algorithms applicable to defense and surveillance autonomy, intelligent agricultural applications, and planetary exploration. Passive perception for world modeling enables stealth operation (since lidars can alert observers) and does not require more expensive or specialized sensors (e.g., radar or lidar). Over the course of this three-phase program, SwRI built components of a vision-only navigation pipeline and tested the result on a vehicle platform in an off-road environment.
Towler, Meera Day, Garza, Harold A., Chambers, David R.
LIPO: Lidar Inertial Odometry for ICP Comparison2025-01-04399/16/2025
We introduce a LiDAR inertial odometry (LIO) framework, called LiPO, that enables direct comparisons of different iterative closest point (ICP) point cloud registration methods. The two common ICP methods we compare are point-to-point (P2P) and point-to-feature (P2F). In our experience, within the context of LIO, P2F-ICP results in less drift and improved mapping accuracy when robots move aggressively through challenging environments when compared to P2P-ICP. However, P2F-ICP methods require more hand-tuned hyper-parameters that make P2F-ICP less general across all environments and motions. In real-world field robotics applications where robots are used across different environments, more general P2P-ICP methods may be preferred despite increased drift. In this paper, we seek to better quantify the trade-off between P2P-ICP and P2F-ICP to help inform when each method should be used. To explore this trade-off, we use LiPO to directly compare ICP methods and test on relevant benchmark datasets as well as on our custom unpiloted ground vehicle (UGV). We find that overall, P2F-ICP has reduced drift and improved mapping accuracy, but, P2P-ICP is more consistent across all environments and motions with minimal drift increase.
Mick, Darwin, Pool, Taylor, Nagaraju, Madankumar Sathenahally, Kaess, Michael, Choset, Howie, Travers, Matthew
A 3D Visibility Label for Mobility Coatings: Enhancing Traffic Safety Through Color and Sensor Perception12-09-02-00097/22/2025
The mobility industry is rapidly advancing towards more autonomous modes of transportation with the adoption of sophisticated self-driving technologies. However, a critical challenge, being the lack of standardized norms for defining, measuring, and ensuring vehicle visibility across various dynamic traffic environments, remains. This lack of awareness of visibility is hindering the development of new regulations for vehicle visibility and the controlled transition to a fully-integrated autonomous future. While current efforts focus on improving sensing technologies like computer vision, LiDAR systems, and sensor fusion development, two key issues remain unresolved: 1 The absence of a representative and realistic three-dimensional color visibility model for measuring and comparing the visibility of complex shapes with large but varying color coated three-dimensional surface areas. 2 The need for enhanced visibility solutions that improve visibility and vehicle detectability for all traffic participants while maintaining color styling freedom. This article presents a new visibility assessment model that measures the three-dimensional Point-of-View (PoV) Visibility of mobility coatings using a 5-parameter Visibility Label. It evaluates how visible a mobility coating color is on a complex three-dimensional vehicle shape from a separate observer’s perspective or an observing device. Key components of this three-dimensional Point-of-View Color Visibility Model include: Moving beyond flat panel color design and measurements to a comprehensive 3D Visibility Model Considering visibility for human vision, computer vision and LiDAR visibility modalities Establishing automotive glossy solid white as reference color The Visibility Label helps users understand how visible complex shapes with different colored coatings are. It quantifies visibility in real-world scenarios, enabling easy comparison between designs and finishes. This simplifies decisions on which coatings enhance visibility most. The article shows that similar colors can have a significant different 3D PoV Color Visibility. It explores ways to enhance this visibility with new color coating technologies, particularly through Crystal Glass Pigment (CGP) formulations that significantly improve 3D color visibility. The example in Appendix B demonstrates an increase in 3D Color Visibility of 200% for Human Visibility and 600% for Full Object Visibility for the addition of 5 weight-percent CGP, and 183% increase for LiDAR Visibility for the addition of 17 weight-percent CGP to an existing silvery metallic automotive refinish color. By emphasizing the importance of 3D PoV Color Visibility in traffic environments where human-controlled, semi- and fully autonomous vehicles coexist, and highlighting the influence of color and shape on vehicle visibility, this comprehensive approach aims to significantly impact road safety today while facilitating a controlled transition to an autonomous future.
Mijnen, Paul W., Moerenburg, Joost H.
U-Shift IV: Innovations and Challenges in Autonomous Urban Mobility2025-01-02847/2/2025
The U-Shift IV represents the latest evolution in modular urban mobility solutions, offering significant advancements over its predecessors. This innovative vehicle concept introduces a distinct separation between the drive module, known as the driveboard, and the transport capsules. The driveboard contains all the necessary components for autonomous driving, allowing it to operate independently. This separation not only enables versatile applications - such as easily swapping capsules for passenger or goods transportation - but also significantly improves the utilization of the driveboard. By allowing a single driveboard to be paired with different capsules, operational efficiency is maximized, enabling continuous deployment of driveboards while the individual capsules are in use. The primary focus of U-Shift IV was to obtain a permit for operating at the Federal Garden Show 2023. To achieve this goal, we built the vehicle around the specific requirements for semi-public road operations which includes narrow streets and pedestrians. This involved integrating necessary modifications across multiple domains, including the e/e-architecture, sensor setup, software stack, and even the design of the driveboard and capsule. By utilizing systematic methods to address regulatory and safety challenges, we ensured that the vehicle met the standards required for autonomous driving in semi-public environments. In this paper, we explore the methodologies employed to achieve regulatory compliance, focusing on sensor integration, software- and e/e-architecture. We discuss our multi-modal sensor setup, which combines camera, lidar and radar to archive redundancy and enhanced environmental perception. Additionally, we provide an overview of our software architecture, emphasizing its role in ensuring safe driving functions and enabling autonomous operations.
Pohl, Eric, Scheibe, Sebastian, Münster, Marco, Osebek, Manuel, Kopp, Gerhard, Siefkes, Tjark
Customized Signalized Intersection Control for Pedestrian Safety Enhancement through Computer Vision Technologies09-13-02-00086/26/2025
The escalating complexity at intersections challenges the safety of the interaction between vehicles and pedestrians, especially for those with mobility impairments. Traditional traffic control systems detect pedestrians through costly technologies such as LiDAR and radar, limiting their adoption due to high costs and static programming. Therefore, the article proposes a customized signalized intersection control (CSIC) algorithm for pedestrian safety enhancement. This algorithm integrates advanced computer vision (CV) algorithms to detect, track, and predict pedestrian movements in real time, enhancing safety at a signalized intersection while remaining economically viable and easily integrated into existing infrastructure. Implemented at a key intersection in Bellevue, the CSIC system achieves a 100% pedestrian passing rate while simultaneously minimizing the average remaining walk time after crossings. The algorithm used in this study demonstrates the potential of combining CV with traditional traffic control mechanisms to create safer urban traffic environments for all road users.
Xia, Rongjing, Fang, Hongchao, Zhang, Chenyang
Reducing LiDAR Cabin Noise with Damped Elastomers a Novel Method for Predicting Performance and Durability2025-01-01145/5/2025
Mechanical light detection and ranging (LiDAR) units utilize spinning lasers to scan surrounding areas to enable limited autonomous driving. The motors within the LiDAR modules create vibration that can propagate through the vehicle frame and become unwanted noise in the cabin of a vehicle. Decoupling the module from the body of the vehicle with highly damped elastomers can reduce the acoustic noise in the cabin and improve the driving experience. Damped elastomers work by absorbing the vibrational energy and dispelling it as low-grade heat. By creating a unique test method to model the behavior of the elastomers, a predictable pattern of the damping ratio yielded insight into the performance of the elastomer throughout the operating temperature range of the LiDAR module. The test method also provides an objective analysis of elastomer durability when exposed to extreme temperatures and loading conditions for extended periods of time. Confidence in elastomer behavior and life span was restored when no signs of performance degradation were present after 30 simulated years of normal loading conditions at extreme temperatures.
Russell, Casey, Masterson, Peter, O'Connell, Kerry
Accuracy of Mobile Phone LiDAR on Crashed Vehicles2025-01-86924/1/2025
Apple’s mobile phone LiDAR capabilities can be used with multiple software applications to capture the geometry of vehicles and smaller objects. The results from different software have been previously researched and compared to traditional ground-based LiDAR. However, results were inconsistent across software applications, with some software being more accurate and others being less accurate. (Technical Paper 2023-01-0614. Miller, Hashemian, Gillihan, Benes.) This paper builds upon existing research by utilizing the updated LiDAR hardware that Apple has added to its iPhone 15 smartphone lineup. This new hardware, in combination with the software application PolyCam, was used to scan a variety of crashed vehicles. These crashed vehicles were also scanned using a FARO 3D scanners and Leica RTC 360 scanners, which have been researched extensively for their accuracy. The PolyCam scans were compared to FARO and Leica scans to determine accuracy for point location and scaling. Previous research has shown that user variability can be minimal (Technical Paper 2023-01-0614. Miller, Hashemian, Gillihan, Benes), so one user utilized the PolyCam application to document real-world crashed vehicles along with 3D scanners for comparative analysis. The authors found that the PolyCam software presented one data set with results similar to the previously researched software application Recon3D, but there were data sets that were less accurate. The authors believe these inconsistencies need to be evaluated further, but believe the accident reconstruction community, and other forensic disciplines, will continue to find accessibility and usability with PolyCam and other iPhone LiDAR applications that are available.
Miller, Seth Higgins, Stogsdill, Michael, McWhirter, Seth
A Step Toward More Accurate 3D Object DetectionTBMG-528154/1/2025
Most 3D object detection methods employ LiDAR sensors to create 3D point clouds of their environment. Simply put, LiDAR sensors use laser beams to rapidly scan and measure the distances of objects and surfaces around the source. However, using LiDAR data alone can lead to errors due to the high sensitivity of LiDAR to noise, especially in adverse weather conditions like during rainfall.
A Motion-Aware Continuous Time LiDAR-Inertial SLAM Framework2025-01-80444/1/2025
Towards the goal of real-time navigation of autonomous robots, the Iterative Closest Point (ICP) based LiDAR odometry methods are a favorable class of Simultaneous Localization and Mapping (SLAM) algorithms for their robustness under any light conditions. However, even with the recent methods, the traditional SLAM challenges persist, where odometry drifts under adversarial conditions such as featureless or dynamic environments, as well as high motion of the robots. In this paper, we present a motion-aware continuous-time LiDAR-inertial SLAM framework. We introduce an efficient EKF-ICP sensor fusion solution by loosely coupling poses from the continuous time ICP and IMU data, designed to improve convergence speed and robustness over existing methods while incorporating a sophisticated motion constraint to maintain accurate localization during rapid motion changes. Our framework is evaluated on the KITTI datasets and artificially motion-induced dataset sequences, demonstrating improvements in SLAM performance in high-motion change environments with loop visualization, making it highly applicable for autonomous navigation in similar high-motion change environments such as uneven terrain and off-road scenarios. We provide various experiments to evaluate quantitatively and qualitatively the estimated trajectories against the ground truth. Our framework ICP-EKF has demonstrated superior trajectory estimation compared to the tightly-coupled EKF SLAM methods FAST-LIO2 and LIO-SAM, and in most instances against the baseline ICP SLAM framework CT-ICP.
Kokenoz, Cigdem, Shaik, Toukheer, Sharma, Abhishek, Pisu, Pierluigi, Li, Bing
Driving in the Rain: Evaluating How Surface Material Properties Affect LiDAR Perception in Autonomous Driving2025-01-80164/1/2025
Light Detection and Ranging (LiDAR) is a promising type of sensor for autonomous driving that utilizes laser technology to provide perceptions and accurate distance measurements of obstacles in the vehicle path. In recent years, there has also been a rise in the implementation of LiDARs in modern and autonomous vehicles to aid self-driving features. However, navigating adverse weather remains one of the biggest challenges in achieving Level 5 full autonomy due to sensor soiling, leading to performance degradation that can pose safety hazards. When driving in rain, raindrops impact the LiDAR sensor assembly and cause attenuation of signals when the light beams undergo reflections and refractions. Consequently, signal detectability, accuracy, and intensity are significantly affected. To date, limited studies have been able to perform objective evaluations of LiDAR performance, most of which faced limitations that hindered realistic, controllable, and repeatable testing. Therefore, this paper reports a fundamental study that employs a previously developed novel wind tunnel testing methodology to investigate and quantify the effects of stress factors affecting LiDAR perception. It was found that soiling characteristics, such as raindrop size distribution and droplet impact kinematics, and material properties, such as surface roughness and wettability, all play critical roles in influencing LiDAR performance to different extents. The results suggest that although LiDAR is an optical type of sensor, its perception does not necessarily align with camera vision, which is closer to human perception. Specifically, hydrophilic surfaces show better advantages over hydrophobic surfaces for LiDAR sensor applications when driving in rain. The overall relationships between material properties, adherent raindrop dynamics, and LiDAR perceptions are summarized.
Pao, Wing Yi, Li, Long, Agelin-Chaab, Martin, Roy, Langis, Knutzen, Julian, Baltazar Y Jimenez, Alexis, Muenker, Klaus, Chakraborty, Anirban, Komar, John
LiDAR Contamination Recognition Based on Optimized PointNet2025-01-80134/1/2025
LiDAR sensors have become an integral component in the realm of autonomous driving, widely utilized in environmental perception and vehicle navigation. However, in real-world road environments, contaminants such as dust and dirt can severely hamper the cleanliness of LiDAR optical windows, thereby degrading operational performance and affecting the overall environmental perception capabilities of intelligent driving systems. Consequently, maintaining the cleanliness of LiDAR optical windows is crucial for sustaining device performance. Unfortunately, the scarcity of publicly available LiDAR contamination datasets poses a challenge to the research and development of contamination identification algorithms. This paper first introduces a method for acquiring LiDAR-pollution datasets. LiDAR data acquisition on urban open roads simulates different types of pollution, including mud and leaves. The constructed dataset meticulously differentiates among the three states with clear labels: no pollution, mud pollution, and leaf pollution. In this paper, the PointNet algorithm is also optimised. The performance of the model is further improved by incorporating the reflection intensity into the model inputs. As a result, the model's accuracy on the test set exceeds 99% and meets the needs of real-time processing. This research not only contributes new ideas and methods for the development of LiDAR optical window contamination algorithms, but also establishes a foundation for future research and applications in LiDAR automatic cleaning.
Wei, Ziyu, Quo, Binyun, Lujia, Ran, Li, Liguang
Hierarchical Feature-Based Localization Using Scene Graphs in Off-Road Navigation2025-01-80464/1/2025
While numerous advancements have been made in autonomous navigation for structured indoor and outdoor environments, these solutions often do not generalize well to off-road settings. There are unique challenges in such settings such as unreliable GPS, limited computational and memory resources, and sparse environmental features, making navigation particularly difficult. In our work, we propose a novel data structure called Hierarchical Dynamic Scene Graphs (HDSG) to address these challenges. HDSG captures environmental information at different resolutions, integrating both geometric and semantic features. It enables various navigation tasks such as localization, loop closure, and human interaction through the visualization of environmental features for remote operators. We evaluated the performance of localizing a robot’s position within the world frame by comparing compact spatial descriptors extracted from semi-consecutive scene graphs, derived from 3D LiDAR point clouds. Compared to directly applying traditional Iterative Closest Point (ICP) algorithms on point clouds, our approach demonstrates that localization on scene graphs is more efficient and accurate. In evaluations using the RELLIS-3D dataset, the HDSG is constructed in at most 5 seconds using only commodity hardware. The overall memory footprint of the HDSG is very compact accounting for only 450-500 MB. Moreover, using the HDSG for robot localization has demonstrated faster and more precise results than using traditional ICP approaches directly on the input point cloud. These results highlight the potential of scene graph-based localization to deliver faster, more memory-efficient, and more accurate performance in unstructured off-road environments, showing a promising foundation for future enhancements and applications.
Alam, Fardifa Fathmiul, Luricich, Federico, Li, Nianyi, Jia, Yunyi, Li, Bing
Decentralized Perception System with Multiple Viewpoints2025-01-80974/1/2025
Vehicle-to-Infrastructure (V2I) cooperation has emerged as a fundamental technology to overcome the limitations of the individual ego-vehicle perception. Onboard perception is limited by the lack of information for understanding the environment, the lack of anticipation, the drop of performance due to occlusions and the physical limitations of embedded sensors. The perception of V2I in a cooperative manner improves the perception range of the ego vehicle by receiving information from the infrastructure that has another point of view, mounted with sensors, such as camera and LiDAR. This technical paper presents a perception pipeline developed for the infrastructure based on images with multiple viewpoints. It is designed to be scalable and has five main components: the image acquisition for the modification of camera settings and to get the pixel data, the object detection for fast and accurate detection of four wheels, two wheels and pedestrians, the data fusion module for robust fusion of the 2D bounding boxes from multiple viewpoints, the object tracking to get the history of movement for each object over time and the generation of perception message for V2I communication. The infrastructure-based solution has been implemented and demonstrated in real-world scenarios, including two different intersections with up to six mounted cameras to cover an extended area. The qualitative results show that the detected objects have high accuracy with similar performances between two different environments, which proves the scalability of the solution. With a not optimized setup for these first deployments, we observe for the whole pipeline an execution time between 226ms and 256ms depending on the number of objects to be fused in the map based on the processing of six cameras.
Picard, Quentin, Morice, Malo, Fadili, Maryem, Pechberti, Steve
A Comparison of UAV LiDAR and Terrestrial Laser Scanner Accuracies2025-01-86944/1/2025
The accident reconstruction community frequently uses Terrestrial LiDAR (TLS) to capture accurate 3D images of vehicle accident sites. This paper compares the accuracy, workflow, benefits, and challenges of Unmanned Aerial Vehicle (UAV) LiDAR, or Airborne Laser Scanning (ALS), to TLS. Two roadways with features relevant to accident reconstruction were selected for testing. ALS missions were conducted at an altitude of 175 feet and a velocity of 4 miles per hour at both sites, followed by 3D scanning using TLS. Survey control points were established to minimize error during cloud-to- cloud TLS registration and to ensure accurate alignment of ALS and TLS point clouds. After data capture, the ALS point cloud was analyzed against the TLS point cloud. Approximately 80% of ALS points were within 1.8 inches of the nearest TLS point, with 64.8% at the rural site and 59.7% at the suburban site within 1.2 inches. These findings indicate that UAV-based LiDAR can achieve comparable accuracy to TLS in accident site documentation, offering potential advantages in efficiency, safety, and accessibility.
Foltz, Steven, Terpstra, Toby, Clarson, Julia
Validating the Sun System in 3ds Max for Recreating Shadows2025-01-86854/1/2025
Shadow positions can be useful in determining the time of day that a photograph was taken and determining the position, size, and orientation of an object casting a shadow in a scene. Astronomical equations can predict the location of the sun relative to the earth, and therefore the position of shadows cast by objects, based on the location’s latitude and longitude as well as the date and time. 3D computer software includes these calculations as a part of their built-in sun systems. In this paper, the authors examine the sun system in the 3D modeling software 3ds Max to determine its accuracy for use in accident reconstruction. A parking lot was scanned using a FARO LiDAR scanner to create a point cloud of the environment. A camera was then set up on a tripod at the environment, and photographs were taken at various times throughout the day from the same location. This environment was 3D modeled in 3ds Max based on the point cloud, and the sun system in 3ds Max was configured using the date and time of the photographs. Photogrammetry techniques were used to align undistorted photographs to the 3D environment, and rendered images were compared to the photographs. The results show that the 3ds Max sun system recreates the sun's position adequately, making it a reliable tool for accurately determining shadow locations and dimensions.
Barreiro, Evan, Erickson, Michael, Smith, Connor, Carter, Neal, Hashemian, Alireza
Development of a Rule-Based Lidar Lane-Keeping Assistance System for Autonomous Vehicles Applications2025-01-80254/1/2025
Lane-keeping is critical for SAE Level 3+ autonomous vehicles, requiring rigorous validation and end-to-end interpretability. All recently U.S.-approved level 3 vehicles are equipped with lidar, likely for accelerating active safety. Lidar offers direct distance measurements, allowing rule-based algorithms compared to camera-based methods, which rely on statistical methods for perception. Furthermore, lidar can support a more comprehensive and detailed approach to studying lane-keeping. This paper proposes a module perceiving oncoming vehicle behavior, as part of a larger behavior-tree structure for adaptive lane-keeping using data from a lidar sensor. The complete behavior tree would include road curvature, speed limits, road types (rural, urban, interstate), and the proximity of objects or humans to lane markings. It also accounts for the lane-keeping behavior, type of adjacent and opposing vehicles, lane occlusion, and weather conditions. The algorithm was evaluated using experimental lidar data collected from driving around Georgia Southern’s campus on one of the behavior tree’s most intensive inputs: oncoming vehicle lane-keeping behavior in two-way, two-lane highways with no physical barriers. Preliminary results include demonstrating one behavior-tree module recognizing an oncoming vehicle’s lane-keeping ability, showing a promising future for interpretable algorithms when using lidar. Existing and novel methods were combined to acquire behavior metrics: Distance to Lane Marking (DTLM), trajectory prediction error (pE), the relative distance between ego- and target vehicles, predicted dividing lane crossings, and the number of vehicle points tracked (NoP).
Soloiu, Valentin, Mehrzed, Shaen, Kroeger, Luke, Pierce, Kody, Sutton, Timothy, Lange, Robin
A Method for Constructing Software and Hardware of a Multi-Source Fusion Intelligent Driving2025-01-71583/19/2025
To meet the requirements of high-precision and stable positioning for autonomous driving vehicles in complex urban environments, this paper designs and develops a multi-sensor fusion intelligent driving hardware and software system based on BDS, IMU, and LiDAR. This system aims to fill the current gap in hardware platform construction and practical verification within multi-sensor fusion technology. Although multi-sensor fusion positioning algorithms have made significant progress in recent years, their application and validation on real hardware platforms remain limited. To address this issue, the system integrates BDS dual antennas, IMU, and LiDAR sensors, enhancing signal reception stability through an optimized layout design and improving hardware structure to accommodate real-time data acquisition and processing in complex environments. The system’s software design is based on factor graph optimization algorithms, which use the global positioning data provided by BDS to constrain the drift of IMU and LiDAR data, ensuring that the system can maintain accurate positioning through IMU and LiDAR collaboration, even when GNSS signals are limited or completely unavailable. Experimental results show that the system’s 3D positioning error in shaded environments is controlled within 7 cm, with a convergence time of no more than 40 seconds. Further statistical analysis reveals a root mean square error (RMSE) of approximately 8 cm and a standard deviation (STD) of 2 cm. During the simulated indoor-outdoor scene transition test, the system’s relative pose error remains stable within 10 cm, demonstrating its adaptability and robustness in diverse and complex scenarios. This study provides a technical reference for the hardware construction and system validation of multi-sensor fusion technology on autonomous driving platforms.
Zhan, KaiDi, Gao, Chengfa, Xu, Dawei, Lan, Minyi, Ding, Rongjing
Background Filtering and Target Detection for Traffic Objects with Roadside LiDAR Point Cloud Data2025-01-71933/19/2025
Roadside perception technology is an essential component of traffic perception technology, primarily relying on various high-performance sensors. Among these, LiDAR stands out as one of the most effective sensors due to its high precision and wide detection range, offering extensive application prospects. This study proposes a voxel density-nearest neighbor background filtering method for roadside LiDAR point cloud data. Firstly, based on the relatively fixed nature of roadside background point clouds, a point cloud filtering method combining voxel density and nearest neighbor is proposed. This method involves voxelizing the point cloud data and using voxel grid density to filter background point clouds, then the results are processed through a neighbor point frame sequence to calculate the average distance of the specified points and compare with a distance threshold to complete accurate background filtering. Secondly, a VGG16-Pointpillars model is proposed, incorporating a CNN network during the point cloud encoding process and adding average pooling weights to enhance point cloud features. The backbone network uses the VGG16 network to extract feature maps of different scales and adds concatenation layers to improve detection accuracy. This method can filter out 99.74% of background point clouds and improve the mean average precision of the target detection model by 3.04%. The model's practical applicability is demonstrated through transfer applications on real-time data.
Liu, Zhiyuan, Rui, Yikang
Multi-Stage Registration and Feature Based Robust and Precise LiDAR SLAM2025-01-71612/21/2025
In a complex and ever-changing environment, achieving stable and precise SLAM (Simultaneous Localization and Mapping) presents a significant challenge. The existing SLAM algorithms often exhibit limitations in design that restrict their performance to specific scenarios; they are prone to failure under conditions of perceptual degradation. SLAM systems should maintain high robustness and accurate state estimation across various environments while minimizing the impact of noise, measurement errors, and external disturbances. This paper proposes a three-stage method for registering LiDAR point cloud. First, the multi-sensor factor graph is combined with historical pose and IMU pre-integration to provide a priori pose estimation; then a new method for extracting planar features is used to describe and filter the local features of the point cloud. Second, the normal distribution transform (NDT) algorithm is used as coarse registration. Third, the feature to feature registration is used for fine registration to achieve iterative optimization of pose. This method also publishes high-frame-rate real-time pose data through IMU, achieving the simultaneous improvement of registration convergence speed, robustness, precision and real-time performance. In addition, this method is suitable for a variety of application scenarios of mechanical and solid-state LiDARs. In terms of experimental verification, this paper uses M2DGR dataset and outdoor testing to verify the precision of the algorithm. The ablation experiments of coarse registration and fine registration are carried out respectively to illustrate the importance of different stages. The results show that the multi-stage registration method is beneficial to improve the precision and robustness of the algorithm.
Li, Zhichao, Tong, Panpan, Shi, Weigang, Bi, Xin
Cryptographic Method for Secure Object Segmentation for Autonomous Driving Perception Systems12-08-01-00082/14/2025
The modern-day vehicle’s driverless or driver-assisted systems are developed by sensing the surroundings using a combination of camera, lidar, and other related sensors by forming an accurate perception of the driving environment. Machine learning algorithms help in forming perception and perform planning and control of the vehicle. The control of the vehicle which reflects safety depends on the accurate understanding of the surroundings by the trained machine learning models by subdividing a camera image fed into multiple segments or objects. The semantic segmentation system comes with the objective of assigning predefined class labels such as tree, road, and the like to each pixel of an image. Any security attacks on pixel classification nodes of the segmentation systems based on deep learning result in the failure of the driver assistance or autonomous vehicle safety functionalities due to a falsely formed perception. The security compromisations on the pixel classification head of the object segmentation systems result in falsely segmented pixels from the incoming camera images by corrupted pixel labels with wrong object classes for the pixels. The popular encoder–decoder-based deep learning object segmentation network is considered, which is vulnerable to these attacks in its last fully connected neural network layer. Hence, the cryptographic solution mechanism is proposed here, where the pixel classes are encrypted and signed in the classification network nodes before applying the activation functions. RSA-512 algorithm-based encryption and DSA-512 algorithm-based digital signature are used to generate the proposed cryptographic components. The added cryptographic components are verified upon segmenting the objects to ensure the segmented object information is free from described security attacks. The performance of the proposed cryptographic secure object segmentation is evaluated for the popular segmentation network called U-Net for the Cityscapes segmentation dataset with the proposed cryptographic algorithms. The performance evaluation indicates that the secure semantic segmentation is performed with satisfactory precision, recall, and F1 scores of 0.86, 0.85, and 0.85, respectively, along with the added security components.
Prashanth, K.Y., Rohitha , U.M.
VTOL UAS Auto-recovery Using a Tested Long-Term Motion Prediction Method to Define the DeckSM_AVTOL_2025-53152/3/2025
Launch, recovery, and deck handling operational performance on smaller ship platforms like Corvettes, Frigates and Destroyers are qualified as the most challenging tasks in the UAS ship-deployment of a VTOL Uncrewed Air System (UAS). One of the main hurdles is the random nature of seaway-created deck motions coupled with ship structure disturbed air wake patterns. The MoD has supported a range of work aimed at bringing Quiescent Period Prediction (QPP) technology to fruition. QPP firstly requires Wave Profiling RADAR to measure the sea wave system out to approximately 2km in the region around a vessel. Secondly these measurements are employed in a wave propagation model to predict the actual wave forces acting on a vessel. Using the wave predictions as inputs to a vessel model makes possible to predict the actual (deterministic as opposed to statistical) motions of a vessel. Wave systems naturally alternate groups of large waves with smaller waves, this property, combined with the predictive ability, allows to identify the quietest (most quiescent) periods in which to conduct wave limited naval operations. Naval mission planners in the Royal Navy, and elsewhere in the World, appreciate the need to maintain rapid, but safe, deck tempo. The fundamental concept is to measure remote sea surface profiles to predict the future wave forces acting upon a vessel. The objective is to expand ship operating deck limits to approximately Sea State 6+. The deck definitions generally empirically measured by using standard rating scales, are replaced by instrumented devices reporting the status of the deck prior to touch-down. In this paper, a thorough discussion describing the QPP deck measuring devices designed to replace piloted cueing is provided. Theory, previous simulation studies and current at-sea testing along with data results, are also discussed. To conclude, the interface of the deck measuring device into the next version of the UK UAS system, is provided. The results of the RADAR trial indicated that the RADAR data was reliable, with the RADAR images matching the physical map. The two-dimensional surface plot showed both the RADAR blocking fence along with an additional target. An additional observation concerning the operation over the deck whilst the ship is experiencing a quiescent ship motion period. The coupled secondary effect documents minimized air wake confusion. This is owing to fewer ship structure excursions into and out of the air flow. To better define deck airflow around the ship the integration of a Doppler LIDAR instrumented federate is proposed. This is meant to predict the future vessel air wake and look for quiescent periods in this paralleling the vessel motion QPP technique.
Ferrier, Bernard, Watson, B., Belmont, M.R., Christmas, J.T.
Adaptive Localization Algorithm for Autonomous Mining Trucks with Motion Distortion Removal2025-01-70271/31/2025
During the operation of autonomous mining trucks in the process of crushing stones, the GPS signal is lost due to signal blockage by the crushing workshop. Simultaneous Localization and Mapping (SLAM) becomes critical for ensuring accurate vehicle positioning and smooth operation. However, the bumpy road conditions and the scarcity of plane and corner feature points in mining environments pose challenges to SLAM algorithms in practical applications, such as pose jumps and insufficient positioning accuracy. To address this, this paper proposes a high-precision positioning algorithm based on inertial navigation 3D signals, incorporating point cloud motion distortion correction, a vehicle roll model, and an Adaptive Kalman Filter (AKF). The goal is to improve the positioning accuracy and stability of autonomous mining trucks in complex scenarios. This paper utilizes real-world operational data from mining vehicles and adopts a 3D point cloud motion distortion correction algorithm to mitigate the impact of bumpy roads on positioning accuracy. Additionally, a dynamic model that considers vehicle sideslip is integrated, and the feedback from the Inertial Measurement Unit (IMU) is fused with the positioning results obtained from LiDAR point cloud registration using Normal Distributions Transform (NDT) through an Adaptive Extended Kalman Filter (AEKF). Furthermore, an error analysis model is designed to enable adaptive adjustment of the algorithm, and the performance of the NDT algorithm is enhanced in open, feature-scarce environments through LiDAR point cloud fusion techniques. Simulation results show that the positioning stability on bumpy roads is improved by approximately 21.2%. The improved algorithm effectively suppresses pose jumps during large turn radii, reducing the average error by 5.94% compared to the traditional Kalman Filter (KF). Moreover, the algorithm demonstrates higher positioning accuracy and stability under sensor failures and adverse weather conditions.
Meng, Chunyang, Song, Kang, Xie, Hui, Xing, Wanyong
Data-Driven Performance Evaluation of Machine Learning for Velocity Estimation Based on Scan Artifacts from LiDAR Sensors12-08-04-00321/21/2025
Light detection and ranging (LiDAR) sensors are increasingly applied to automated driving vehicles. Microelectromechanical systems are an established technology for making LiDAR sensors cost-effective and mechanically robust for automotive applications. These sensors scan their environment using a pulsed laser to record a point cloud. The scanning process leads in the point cloud to a distortion of objects with a relative velocity to the sensor. The consecutive generation and processing of points offers the opportunity to enrich the measured object data from the LiDAR sensors with velocity information by extracting information with the help of machine learning, without the need for object tracking. Turning it into a so-called 4D-LiDAR. This allows object detection, object tracking, and sensor data fusion based on LiDAR sensor data to be optimized. Moreover, this affects all overlying levels of autonomous driving functions or advanced driver assistance systems. However, since such sensor-specific effects are rarely available in public datasets and the velocities of target objects are not included as ground truth in these datasets, it makes sense to enrich the limited real-world data with synthetic data. Therefore, this article discusses how such datasets can be created and combined to efficiently estimate velocities on real-world data using the novel method named VeloPoints.
Haas, Lukas, Haider, Arsalan, Kastner, Ludwig, Kuba, Matthias, Zeh, Thomas, Jakobi, Martin, Koch, Alexander Walter
SAE TOMORROW TODAY: How AI Alignment Makes AVs Safer134951/10/2025
In order for AVs to perform safely and reliably, we need to teach them the language of human preference and expectations--and accelerating AI alignment can do just that. Enter Kognic, the industry-leading annotation platform for sensor-fusion datasets (e.g., camera, radar, and LIDAR data). By helping companies gather, organize, and refine massive datasets used for training AI models, Kognic is helping to ensure that AD/ADAS perform reliably and meet safety standards--all while minimizing costs and optimizing teams. To learn more, we sat down with Daniel Langkilde, Co-Founder and CEO, to discuss why the future of autonomous driving depends on effectively managing AI-driven datasets and how Kognic is leading dataset management for safety-critical AI. 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, Twitter, and YouTube. Follow host Grayson Brulte on LinkedIn, Twitter, and Instagram.
Hineman, Marcie
Unified Multi-Modal Multi-Agent Cooperative Perception Framework for Intelligent Transportation Systems2024-01-702812/13/2024
Cooperative perception has attracted wide attention given its capability to leverage shared information across connected automated vehicles (CAVs) and smart infrastructure to address the occlusion and sensing range limitation issues. To date, existing research is mainly focused on prototyping cooperative perception using only one type of sensor such as LiDAR and camera. In such cases, the performance of cooperative perception is constrained by individual sensor limitations. To exploit the multi-modality of sensors to further improve distant object detection accuracy, in this paper, we propose a unified multi-modal multi-agent cooperative perception framework that integrates camera and LiDAR data to enhance perception performance in intelligent transportation systems. By leveraging the complementary strengths of LiDAR and camera sensors, our framework utilizes the geometry information from LiDAR and the semantic information from cameras to achieve an accurate cooperative perception system. In order to fuse the multi-agent and multi-modal features, we use a bird’s-eye view (BEV) space as the consistent and unified feature representations and employ a transformer-based network for effective multi-agent multi-modal BEV feature fusion. We validate our method on the OPV2V and V2XSim benchmarks, achieving state-of-the-art performance in 3D cooperative perception tasks. The proposed framework significantly improves object detection accuracy and robustness, especially in complex traffic scenarios with occlusions such as dense intersections.
Meng, Zonglin, Xia, Xin, Zheng, Zhaoliang, Gao, Letian, Liu, Wei, Zhu, Jiaqi, Ma, Jiaqi
AI-Based Smart Vehicle Parking System2024-01-524212/10/2024
This project presents the development of an advanced Autonomous Mobile Robot (AMR) designed to autonomously lift and maneuver four-wheel drive vehicles into parking spaces without human intervention. By leveraging cutting-edge camera and sensor technologies, the AMR integrates LIDAR for precise distance measurements and obstacle detection, high-resolution cameras for capturing detailed images of the parking environment, and object recognition algorithms for accurately identifying and selecting available parking spaces. These integrated technologies enable the AMR to navigate complex parking lots, optimize space utilization, and provide seamless automated parking. The AMR autonomously detects free parking spaces, lifts the vehicle, and parks it with high precision, making the entire parking process autonomous and highly efficient. This project pushes the boundaries of autonomous vehicle technology, aiming to contribute significantly to smarter and more efficient urban mobility systems.
Atheef, M. Syed, Sundar, K. Sham, Kumar, P. P. Prem, Karthika, J.
Autonomous Mobile Robot for Gas Leak Localization2024-01-522612/10/2024
LIDAR-based autonomous mobile robots (AMRs) are gradually being used for gas detection in industries. They detect tiny changes in the composition of the environment in indoor areas that is too risky for humans, making it ideal for the detection of gases. This current work focusses on the basic aspect of gas detection and avoiding unwanted accidents in industrial sectors by using an AMR with LIDAR sensor capable of autonomous navigation and MQ2 a gas detection sensor for identifying the leakages including toxic and explosive gases, and can alert the necessary personnel in real-time by using simultaneous localization and mapping (SLAM) algorithm and gas distribution mapping (GDM). GDM in accordance with SLAM algorithm directs the robot towards the leakage point immediately thereby avoiding accidents. Raspberry Pi 4 is used for efficient data processing and hardware part accomplished with PGM45775 DC motor for movements with 2D LIDAR allowing 360° mapping. The adoption of LIDAR-based AMRs for gas detection is expected to increase in the future, as more industries realize the benefits of this technology.
Feroz Ali, L., Madhankumar, S., Hariush, V.C., Jahath Pranav, R., Jayadeep, J., Jeffrey, S.
Best Practices in Sensor Selection for Object Detection in Autonomous Driving: A Practitioner’s Perspective2024-28-021812/5/2024
Object detection (OD) is one of the most important aspects in Autonomous Driving (AD) application. This depends on the strategic sensor’s selection and placement of sensors around the vehicle. The sensors should be selected based on various constraints such as range, use-case, and cost limitation. This paper introduces a systematic approach for identifying the optimal practices for selecting sensors in AD object detection, offering guidance for those looking to expand their expertise in this field and select the most suitable sensors accordingly. In general, object detection typically involves utilizing RADAR, LiDAR, and cameras. RADAR excels in accurately measuring longitudinal distances over both long and short ranges, but its accuracy in lateral distances is limited. LiDAR is known for its ability to provide accurate range data, but it struggles to identify objects in various weather conditions. On the other hand, camera-based systems offer superior recognition capabilities but lack the precision in range resolution. Fusion of all the three sensors could improve object detection results, however at a higher cost and may be redundant in some cases. In autonomous driving, different functions like dynamic fusion, static fusion, and road model are used to detect a variety of objects like vehicles, motorcycles, guardrails, and road lanes. The paper presents an in-depth analysis of the mechanisms of each sensor, the nature of the data it generates, its level of accuracy, and the limitations it encounters in detecting various objects. For each object, the paper outlines important steps and recommendations that can be implemented to achieve optimal results. This paper elucidates a framework for multi-sensor fusion in object detection, demonstrating superior performance through a practical use case. Model output is rigorously validated against ground truth data using proven devices. The proposed methodology yields demonstrably improved and refined obstacle and environment classification.
Maktedar, Asrarulhaq, Chatterjee, Mayurika
INTEGRATING SENSOR DATA: Selecting an ADAS decision-making process24AUTP12_0112/1/2024
Exactly when sensor fusion occurs in ADAS operations, late or early, impacts the entire system. Governments have been studying Advanced Driver Assistance Systems (ADAS) since at least the late 1980s. Europe's Generic Intelligent Driver Support initiative ran from 1989 to 1992 and aimed “to determine the requirements and design standards for a class of intelligent driver support systems which will conform with the information requirements and performance capabilities of the individual drivers.” Automakers have spent the past 30 years rolling out such systems to the buying public. Toyota and Mitsubishi started offering radar-based cruise control to Japanese drivers in the mid-1990s. Mercedes-Benz took the technology global with its Distronic adaptive cruise control in the 1998 S-Class. Cadillac followed that two years later with FLIR-based night vision on the 2000 Deville DTS. And in 2003, Toyota launched an automated parallel parking technology called Intelligent Parking Assist on the Prius.
Ramsey, Jonathon
Research on the Application of High-Static-Low-Dynamic Stiffness Vibration Isolator in Electro-Optical Pods of Unmanned Aerial Vehicles2024-01-701511/15/2024
With the rapid advancement in unmanned aerial vehicle (UAV) technology, the demand for stable and high-precision electro-optical (EO) pods, such as cameras, lidar sensors, and infrared imaging systems, has significantly increased. However, the inherent vibrations generated by the UAV’s propulsion system and aerodynamic disturbances pose significant challenges to the stability and accuracy of these payloads. To address this issue, this paper presents a study on the application of high-static low-dynamic stiffness (HSLDS) vibration isolation devices in EO payloads mounted on UAVs. The HSLDS system is designed to effectively isolate low-frequency and high-amplitude vibrations while maintaining high static stiffness, ensuring both stability during hovering and precise pointing capabilities. A nonlinear dynamic system model with two degrees of freedom is formulated for an EO pod supported by HSLDS isolators at both ends. The model’s natural frequencies are determined, and approximate expressions for the displacement transmissibility at both ends are derived. These expressions facilitate a comparative analysis of the isolation performance between the HSLDS-supported system and a system with linear elastic supports. Furthermore, the influence of system parameters on isolation performance is exhaustively investigated and summarized. The findings reveal that the HSLDS-supported system exhibits a broader effective frequency range for vibration isolation and improved isolation performance at operating frequencies, as compared to a system utilizing linear elastic supports under identical base excitation conditions. This research underscores the potential of HSLDS devices as an effective solution for enhancing the stability and accuracy of EO pods mounted on UAVs, thereby advancing the capabilities of these aerial platforms in critical applications such as reconnaissance, surveillance, and target tracking.
Tian, Yishen, Guo, Gaofeng, Wang, Guangzhao, Wei, Wan, Bao, Lingcong, Dong, Guan, Li, Liujie
Autonomous Landing of UAV in Non-Cooperative Environments Based on Fusion of Monocular and Stereo Depth Estimation2024-01-701911/15/2024
In non-cooperative environments, unmanned aerial vehicles (UAVs) have to land without artificial markers, which is a key step towards achieving full autonomy. However, the existing vision-based schemes have the common problems of poor robustness and generalization, and the LiDAR-based schemes have the disadvantages of low resolution, high power consumption and high weight. In this paper, we propose an UAV landing system equipped with a binocular camera to preform 3D reconstruction and select the safe landing zone. The whole system only consists of a stereo camera, and the innovation of the solution is fusing the stereo matching algorithm and monocular depth estimation(MDE) model to get a robust prediction on the metric depth. The whole landing system consists of a stereo matching module, a monocular depth estimation (MDE) module, a depth fusion module, and a safe landing zone selection module. The stereo matching module uses Semi-Global Matching (SGM) algorithm to calculate the binocular disparities to get the dense metric depth of each pixel, and is deployed on GPU to meet the real-time requirements. The MDE module conducts relative depth estima-tion on the left-eye image, and is also deployed on GPU to improve the inference speed. The output of the MDE module has the advantages of high accuracy and excellent generalization but it is ambiguous in scale, and that of the stereo matching module is deterministic in scale but susceptible to illumination and moving objects. Considering the above features, the depth fusion module fuses the relative depth estimation result and the metric depth information into a robust and accurate metric depth map. With the metric depth map and camera intrinsic parameters, the safe landing zone selection module calculates the first and second order derivatives to detect the obstacles and finds the safe landing zone. In AirSim, we build a UAV Hardware-In-the-Loop (HIL) simulation system, and carry out a series of autonomous landing experiments. The results show that our landing scheme performs 3D reconstruction of the landing terrain and selects safe landing zone with high efficiency and reliability.
Zhou, YiBiao, Zhang, BiHui
Why Honda Wants FMCW LiDAR in Next Generation MobilityTBMG-5191911/1/2024
In September, after several months of evaluating the market, “Honda Xcelerator Ventures” — the automotive manufacturer’s startup investment subsidiary — made a major investment award to California-based silicon photonics startup SiLC Technologies, Inc., to develop next generation Frequency-Modulated Continuous Wave (FMCW) LiDAR for “all types of mobility.”
Towards Automation of Reference Data Generation for ADAS/AD Functions Development – ALiVA Framework2024-28-002210/17/2024
The advancements towards autonomous driving have propelled the need for reference/ground truth data for development and validation of various functionalities. Traditional data labelling methods are time consuming, skills intensive and have many drawbacks. These challenges are addressed through ALiVA (automatic lidar, image & video annotator), a semi-automated framework assisting for event detection and generation of reference data through annotation/labelling of video & point-cloud data. ALiVA is capable of processing large volumes of camera & lidar sensor data. Main pillars of framework are object detection-classification models, object tracking algorithms, cognitive algorithms and annotation results review functionality. Automatic object detection functionality creates a precise bounding box around the area of interest and assigns class labels to annotated objects. Object tracking algorithms tracks detected objects in video frames, provides a unique object id for each object and performs distance ranging. A unique feature of cognitive algorithms is the elimination of non-realistic objects of interests which appear in billboards or advertisements on buses/trucks. The framework also has a feature of event detection like overtaking scenarios or pedestrians/animals crossing the roads. Annotation review functionality is provided where assessment and correction of auto annotated data can be done manually. The results can be saved in standard file formats such as txt, csv, Json and open ASAM, ensuring compatibility across different systems. ALiVA replaces traditional annotation methods, thereby reducing the effort, the need for skilled resources and the time required to annotate large datasets. This eliminates human biases, manual errors and inconsistencies. ALiVA is validated for numerous customer requirements and offers a large amount and variety of data to quantify the benefits offered. Some of the distinguishing features are models and functionalities that are optimized for Asian road scenarios, which are typically characterized by very high road density. It is platform independent, adaptable to newer requirements, complements newer event definitions for data segmentation and works both in cloud environments for Data as a service and as a standalone desktop application.
Mardhekar, Amogh, Pawar, Rushikesh, Mohod, Rucha, Shirudkar, Rohit, Hivarkar, Umesh N.
A Powerful Vision-Based Autonomy Alternative to LiDAR, Radar, GPSTBMG-5174710/1/2024
Southwest Research Institute has developed off-road autonomous driving tools with a focus on stealth for the military and agility for space and agriculture clients. The vision-based system pairs stereo cameras with novel algorithms, eliminating the need for LiDAR and active sensors.
Approach for Extrinsic Calibration of a Light Detection and Ranging Sensor and a Monocular Camera Using Bounding Boxes2024-01-50929/4/2024
Sensor calibration plays an important role in determining overall navigation accuracy of an autonomous vehicle (AV). Calibrating the AV’s perception sensors, typically, involves placing a prominent object in a region visible to the sensors and then taking measurements to further analyses. The analysis involves developing a mathematical model that relates the AV’s perception sensors using the measurements taken of the prominent object. The calibration process has multiple steps that require high precision, which tend to be tedious and time-consuming. Worse, calibration has to be repeated to determine new extrinsic parameters whenever either one of the sensors move. Extrinsic calibration approaches for LiDAR and camera depend on objects or landmarks with distinct features, like hard edges or large planar faces that are easy to identify in measurements. The current work proposes a method for extrinsically calibrating a LiDAR and a forward-facing monocular camera using 3D and 2D bounding boxes. The proposed algorithm was tested using the KITTI dataset and experimental data. The rotation matrix is evaluated by calculating its Euler angles and comparing them to the ideal Euler angles that describe the ideal angular orientation of the LiDAR with respect to the camera. The comparison shows that calibration algorithm’s rotation matrix is approximately close to both the ideal and the KITTI dataset rotation matrices. The corresponding translation vector is shown to be close to expected values as well. The results from the experimental data were evaluated and verified by projecting cluster measurements of the prominent objects on to corresponding images.
Omwansa, Mark, Sharma, Sachin, Meyer, Richard, Brown, Nicholas
Items per page:
1 – 50 of 444