PaperlayerView exact revision on arXiv

Source-linked AI summary

Perception and Sensing for Autonomous Vehicles Under Adverse Weather Conditions: A Survey

Yuxiao Zhang, Alexander Carballo, Hanting Yang, Kazuya Takeda

arXiv:2112.08936v2cs.RO

TL;DR

Adverse weather degrades ADS sensing and remains a barrier to level 4 or higher autonomy. This survey analyzes weather effects and synthesizes hardware, software, validation, auxiliary, and future-sensor approaches, concluding that progress is stronger for rain and fog than snow while robust fusion, sophisticated networks, and V2X/IoT are prominent directions.

  • Problem

    Adverse weather impairs ADS perception and vehicle operation, while systematic coverage of weather phenomena, sensor impacts, and available solutions has been limited.

  • Method

    The paper surveys weather effects on ADS sensors and components, analyzes them statistically, and reviews hardware, software, experimental, auxiliary, dataset, simulator, and facility-based solutions.

  • Results

    Rain and fog performance has improved with advances in test instruments, LiDAR architectures, and computer vision, whereas snow remains focused on dataset expansion and perception enhancement.

  • Takeaways & Limitations

    The paper identifies robust sensor fusion, sophisticated networks and computer-vision models, and V2X and IoT support as major research directions for adverse-weather driving.

  • Takeaways & Limitations

    Hardware limitations involving temperature and humidity endurance remain barely analyzed systematically, and LiDAR and other sensor weaknesses persist in adverse conditions.

Abstract

from arXiv · show

Automated Driving Systems (ADS) open up a new domain for the automotive industry and offer new possibilities for future transportation with higher efficiency and comfortable experiences. However, autonomous driving under adverse weather conditions has been the problem that keeps autonomous vehicles (AVs) from going to level 4 or higher autonomy for a long time. This paper assesses the influences and challenges that weather brings to ADS sensors in an analytic and statistical way, and surveys the solutions against inclement weather conditions. State-of-the-art techniques on perception enhancement with regard to each kind of weather are thoroughly reported. External auxiliary solutions, weather conditions coverage in currently available datasets, simulators, and experimental facilities with weather chambers are distinctly sorted out. Additionally, potential future ADS sensors candidates and approaches beyond common senses are provided. By looking into all kinds of major weather problems the autonomous driving field is currently facing, and reviewing both hardware and computer science solutions in recent years, this survey points out the main moving trends of adverse weather problems in autonomous driving, i.e., advanced sensor fusions, more sophisticated networks, and V2X & IoT technologies; and also the limitations brought by emerging 1550 nm LiDARs. In general, this work contributes a holistic overview of the obstacles and directions of ADS development in terms of adverse weather driving conditions.

1. Introduction

Adverse weather remains a major barrier to higher-level autonomous driving because it degrades sensing, vehicle operation, and safety. This survey addresses the gap through broad analysis of weather impacts and reported hardware, software, validation, and future-direction solutions.

  • Motivation: Current autonomous vehicles barely operate during heavy rain or snow because of safety issues and operational constraints.The Mcity shuttle shuts down when windshield wipers must run continuously, while the Sohjoa Baltic autonomous minibus experienced overnight charging failure in the Estonian winter.
  • Research gap: Existing literature has addressed individual sensors or selected weather conditions, but lacks coverage of all weather phenomena, impacts, components, and hardware-software solutions.The survey identifies limited attention to snow compared with rain and fog and presents itself as filling this broader literature gap.
  • Contributions: The paper statistically analyzes direct and secondary weather influences on ADS sensors and thoroughly reports sensor-fusion, mechanical, perception-enhancement, and related technologies.It also provides a quick index to the corresponding literature.
  • Contributions: The survey conducts experimental validations of perception-enhancement solutions and proposes trends and future research directions, including discussion of limitations facing autonomous driving.The paper organizes these topics across weather influences, sensing solutions, perception methods, auxiliary approaches, and future research.

2. Overview of Autonomous Vehicles

Autonomous vehicles use multiple sensing and positioning modalities to perceive objects, localize themselves, track motion, and support safe planning. Adverse weather changes environmental and vehicle states, impairing these functions and motivating complementary sensor configurations and positioning methods.

  • ADS architecture: Weather alters environmental states and can affect vehicle states through wind and road-surface changes, increasing difficulty in detection, tracking, localization, planning, and control.Changes in ego-vehicle states can further influence sensor operation and vehicle behavior.
  • LiDAR: LiDAR provides accurate, illumination-independent 3D sensing, with key attributes including range, accuracy, point density, scan speed, wavelength, robustness, form factor, and cost.Modern LiDARs may switch between strongest and last signal returns; in denser fog, the last return can overlap more closely with a reference point cloud.
  • Cameras: Cameras remain indispensable in ADS but are among the most vulnerable sensors under adverse weather conditions.The overview places cameras alongside LiDAR, radar, ultrasonic sensing, and external positioning systems as complementary modalities.
  • Radar: Automotive radar measures object distance, speed, and direction using reflected radio waves, commonly at 24–77 GHz or, for some on-chip systems, 122 GHz.FMCW radar continuously varies transmitted frequency and offers speed measurement together with improved range resolution and accuracy.
  • Localization: Localization combines methods such as GNSS, non-GNSS radio, odometry, and inertial navigation, often using sensor fusion, Kalman filtering, and motion models to improve accuracy and reduce drift.One cited GNSS/INS system reports 8 mm horizontal position accuracy and about 0.005° roll/pitch accuracy.
  • Perception functions: ADS perception extracts objects and their locations, while localization estimates ego-position and tracking predicts other parties’ trajectories for subsequent planning and control.The overview distinguishes object detection, localization, and tracking as linked but separate perception functions.

3. Adverse Weather Influences

Adverse weather degrades sensing through attenuation, noise, occlusion, distortion, and environmental contamination, with effects varying substantially by sensor and condition. Heavy precipitation, snow, fog, strong light, and reflective or obstructed surfaces can impair perception and downstream autonomous-driving functions.

  • Rain and fog: Heavy rain produces agglomerate fog and fake LiDAR obstacles, while signal reflection intensity drops significantly at rain rates of 40 and 95 mm/hr.The survey treats heavy rain similarly to dense fog or smoke when assessing its effects.
  • Snow: Snow swirl created front and rear point-cloud voids for an Ouster OS1-64 LiDAR, while similar voids also appeared without a preceding vehicle.Possible causes include heavy snowfall swirl, snowflake accumulation on the optical window, and melted snow droplets.
  • Radar: Radar attenuated about 10 dB/km at 77 GHz in 25 mm/hr rain, compared with about 35 dB/km for 905 nm LiDAR under visibility below 0.5 km.Radar remained more weather-resilient, although low spatial resolution limited pedestrian detection and object shape and size classification.
  • Cameras and light: Cameras can lose or distort input from a single water drop, fog, strong light, or reflections, with night or fog conditions producing up to a 40% rise in miss rate.Extreme illumination can reduce camera visibility to almost zero, and LiDAR can also show a large black area around a strong light source.
  • Contamination and damage: Unknown debris and surface contamination can block sensors, while near-homogeneous dust particles reduced LiDAR maximum range by 75%.Weather-related damage can also disrupt signal reception, images, and electronic-device integrity.

4. Sensor Fusion and Mechanical Solutions

Sensor fusion and mechanical interventions are presented as complementary responses to adverse weather, combining heterogeneous sensing with protection, cleaning, coating, and recalibration measures. Robustness depends on both sensor diversity and weather-specific weighting, while hardware durability and alignment remain practical constraints.

  • 4.1. Sensor fusion modalities: Traditional LiDAR- or camera-only architectures are insufficient for safe adverse-weather navigation, motivating multimodal sensor fusion.
  • 4.1. Sensor fusion modalities: LiDAR-camera, radar-LiDAR, radar-camera, infrared, gated-camera, stereo-camera, and weather-sensor combinations improve complementary perception capabilities under difficult conditions.Radar can support long-distance detection and velocity estimation, while cameras contribute target categorization; a fog-penetrating radar also supplemented LiDAR for SLAM.
  • 4.1.2. Camera dominant: A fused thermal LWIR, radar, and visible-camera AEB suite avoided mannequin collisions in adverse scenarios where normal AEB almost always collided.The suite was evaluated during daytime, nighttime, tunnel exits into sun glare, and thick fog; LWIR spans 8 μm to 14 μm.
  • 4.1.3. Comprehensive fusions: Sensor diversity raises the lower bound of perception robustness, but weather-specific sensor weighting and quantified parameters determine the achievable reliability ceiling.
  • 4.2. Mechanical solutions: Mechanical solutions include ultrasonic cleaning, hydrophobic coatings, low-reflectance dashboard materials, and protective housings to reduce contamination, condensation, and reflections.Hydrophobic membranes reportedly improve visual distance and decrease minimum visual angle by almost 34%, while flock material has visible-light and NIR reflectance below 0.5%.
  • 4.2.3. Integrity and re-calibration: Sensor housings and mounting structures require integrity monitoring and recalibration because lifetime effects, deformation, and vehicle-state changes can shift sensor alignment.

5. Perception Enhancement Methods and Experimental Validations in Different Weather Conditions

Adverse weather degrades ADS perception, so the survey reviews weather-specific hardware, sensor-fusion, and learning-based methods for restoring reliable detection. It covers rain and fog effects, experimental validations, and limitations of current enhancement techniques.

  • Weather-induced attenuation and noise impair ADS sensors, making object detection the central perception-enhancement challenge in adverse conditions.The survey organizes perception methods by weather type after discussing mechanical protection and navigation techniques.
  • Rain: Rain studies quantify LiDAR degradation through changes in range, signal intensity, and detected-point counts, while threshold-based countermeasures require precise benchmarks.Reported evaluations include road signs, building facades, and asphalt, but the stationary setup and surface orientations limit impartiality.
  • Rain: Camera de-raining methods address falling and adherent raindrops using learned streak-removal and generative approaches trained across rain intensities.One network used HOG and auto-correlation loss for orientation consistency and repetitive-streak suppression, with real-data validation reported as superior.
  • Rain: Rain-removal algorithms remain challenged by adherent drops, heavy rain, and dynamic scenes, while real-time adherent-drop removal introduces processing-latency trade-offs.PSNR and SSIM are widely used but are described as less promising than deep-learning and CNN approaches for raindrop removal.
  • Fog in point clouds: Fog-oriented LiDAR methods use visibility-range classification, multiple echoes, full waveforms, and point-cloud denoising to reject weather artifacts and recover targets.HDDM+ recognizes waveform differences among fog, rain, dust, snow, leaves, and fences; WeatherNet distinguishes fog- or rain-induced point-cloud clusters.
  • Fog in images: Multimodal fusion combines LiDAR and camera features to improve fog-condition detection beyond LiDAR alone, while de-hazing methods recover scene radiance from hazy images.GAN-based unpaired translation avoids one-to-one source-target correspondence, and haze removal increased valid detections, especially for partially obscured objects.
  • Fog in images: Fog and haze enhancement remains constrained by scarce real paired ground truth, reliance on synthetic training data, and the ill-posed nature of image de-hazing.Existing quality assessment is therefore often limited to non-reference image-quality metrics.

5.3. Snow

Snow degrades camera, LiDAR, and radar perception through obscured paths, reduced clarity, and point-cloud noise. Reviewed countermeasures include multimodal fusion, synthetic snow generation, and dynamic filtering, although dense snowfall and scene-dependent thresholds remain difficult.

  • Snow pathfinding: Snow-covered scenes motivate drivable-path detection using RGB filtering, thermal imagery, and camera–LiDAR–radar fusion.Thermal sensing helps represent pedestrians and objects that RGB may miss, while multimodal encoders account for unequal sensor data densities.
  • Point-cloud de-snowing: Dynamic radius outlier removal adapts each point’s search radius to geometry, preserving distant structure while removing near-sensor snow noise.The reported precision improvement is nearly 4 times that of normal ROR filters, with essential points preserved from 6 m to 18 m.
  • Point-cloud de-snowing: LIOR filters snow using intensity differences from real objects and can retain more environmental points than DROR under suitable optical conditions.Its threshold is targeted to subject optical properties rather than relying only on spatial structure.
  • Image de-snowing: Camera de-snowing uses deep dense multiscale networks, while physically based rendering synthesizes snowflakes with depth, motion blur, and meteorological properties.Synthetic snow can be generated from reconstructed 3D scenes using stereo or simulator-derived depth.
  • Point-cloud de-snowing: DSOR combines statistical and dynamic range-based filtering, matching or exceeding DROR for denoising and feature preservation while computing faster.Its threshold changes with range, and the spacing factor controls filter mildness.
  • Point-cloud de-snowing: Intensity and entropy filters remain scene-sensitive: intensity filtering requires an exact interval, while entropy filtering struggles in very dense snowfall.The tested intensity interval was [0.03, 0.15], and the entropy calculation used a local radius of 0.25 m with solitary points having fewer than 15 neighbors penalized.

5.4. Light related

Strong illumination can severely impair RGB cameras and even LiDAR, whereas thermal imaging retains useful object information. Enhancement methods and thermal-based classification partially restore perception under these conditions.

  • Light-related degradation: Strong light can reduce camera visibility nearly to zero and produce LiDAR black regions around the source, while glossy reflections complicate exposure.The paper identifies HDR imaging and reflection-handling methods as responses to difficult illumination.
  • Thermal imaging: At 40 m from a Xenon source reaching 200 klx, RGB and LiDAR detected too few points for recognition, but thermal imaging still distinguished a board beneath the source.The experiment reports four mannequin points, three reflective-target points, and ten black-vehicle points.
  • Thermal enhancement: Multiscale Retinex restores color and contrast, while a bio-inspired retina removes illumination variation and enhances static and dynamic contours.Both methods were applied to thermal images before object-classification evaluation.
  • Thermal validation: YOLO3 recall was zero for RGB, partial for unenhanced thermal, and substantially higher after either thermal enhancement method.Bio-inspired retina achieved recalls of 67.1%, 45%, and 78.6% for person, car 1, and car 2, respectively.

5.5. Contamination

Contamination can invade sensor sightlines and sharply challenge ADS robustness. Countermeasures include soiling datasets, automated cleaning controls, and contamination-resistant laser scanners.

  • Contamination effects: Mud and other contamination can severely impair sensor perception by obstructing the line of sight.The paper illustrates this effect on a Cadillac XT5 backup camera after off-road driving in rain.
  • Detection and datasets: SoilingNet covers opaque and transparent camera soiling and supports GAN-based augmentation for lens-soiling detection.This addresses contamination as a distinct condition from rain or snow.
  • Mechanical solutions: Windshield cleaning can be automated by distinguishing liquid and solid contamination through total internal reflection and red-light intensity distributions.The method controls both wiper and nozzle operation.
  • Hardware resilience: SICK’s microScan 3 uses 845 nm lasers and HDDM+ to detect nearby objects despite severe contamination such as sawdust on its emitter window.The scanner covers a 275° angle and is designed for outdoor weather conditions.

6. Classification and Assessment

Weather-aware ADS requires classification of weather, visibility, and risk rather than perception enhancement alone. The surveyed methods range from image and LiDAR classifiers to multimodal and infrastructure-assisted approaches, but real-time visibility and risk assessment remain open challenges.

  • Weather classification: Weather classification has progressed from few-class image judgments to finer categories, yet traditional image classification remains unsaturated and may need multimodal inputs.The survey emphasizes precise classification for targeted ADS decisions.
  • Weather classification: Multi-echo LiDAR classifies rain and fog using secondary echoes and statistical properties of echo distances.The method transforms point clouds into grid matrices before classification.
  • Infrastructure assistance: Roadside cameras and future IoT or V2X systems could extend visibility classification beyond onboard sensing, but mobile real-time capability remains unclear.Conditions beyond fog may require sophisticated sensor fusion.
  • Visibility classification: Visibility estimation methods combine image features, Shannon entropy, convolutional layers, and classifiers to predict discrete visibility ranges.One hybrid network outputs 0–50 m, 50–150 m, and above 150 m classes.
  • Visibility classification: The hybrid visibility estimator achieved overall accuracy of 85% or higher, enabling ADS to establish dangerous-low-visibility thresholds.Its training used BOSCH range-finder ground truth and public synthetic benchmarking with FORSI.
  • Risk assessment: Visibility-based safety rules must account for impaired sensors and severe weather, motivating reliability analysis within decision-making and control.The survey notes that autonomous-driving literature rarely focuses specifically on risk assessment under inclement weather.

7. Localization and Mapping

Adverse weather degrades localization through unreliable positioning, changing visual and LiDAR landmarks, and lateral-reference confusion. The surveyed responses combine robust maps, alternative sensing, sensor fusion, and profile reconstruction.

  • Localization robustness: GPS or INS alone can have over 1 meter double-standard deviation on most roads, while WhONet reduced positioning error by over 90% in GNSS-deprived wet and muddy-road tests.The authors caution that weather-specific performance remains unverified.
  • SLAM: Seasonal appearance changes, foliage, and snow-covered ground compromise SLAM feature descriptors, while drift and incomplete maps limit long-distance outdoor navigation.These limitations make SLAM less competitive than pre-built-map localization for autonomous driving.
  • Map-based methods: Alternative localization approaches include LGPR ground-reflection matching and all-weather surfel maps built from year-long LiDAR data.LGPR uses under-vehicle electromagnetic reflections, while surfel maps model season-invariant surface patches.
  • Sensor solutions: Radar-only localization and multisensory place recognition improve robustness in rain and snow, while Gaussian-mixture LiDAR maps reduce snowy lateral and longitudinal RSM error by around 80%.The Gaussian-mixture representation compresses point clouds into parametric 2.5D maps.
  • Lateral localization: PCA-based LiDAR edge-profile matching reconstructed missing road information and reduced lateral movement error to 15 cm with 96.4% localization accuracy in critical environments.The method suppresses false lane lines caused by snow, wet surfaces, road wear, and vegetation.

8. Planning and Control

Weather changes route conditions, vehicle dynamics, control difficulty, and passenger comfort requirements. The surveyed approaches emphasize rerouting, uncertainty-aware behavior, adaptive safety constraints, and stabilization under degraded visibility or traction.

  • Global planning: Extreme weather and related disasters can damage roads or threaten outdoor activities, requiring global route planners to navigate AVs around affected areas.The passage frames this as a long-distance field-testing challenge because such conditions are unpredictable.
  • Local planning: Repeatedly driving identical lines can create centimeter-scale route precision while producing worn grooves or slippery winter tracks.This creates a local route-planning and road-wearing prevention concern.
  • Motion planning: Uncertainty-aware planning under limited visibility imitates cautious human behavior by lowering speed, preparing to yield, and considering noncompliant road users.The planner uses worst-case assumptions to promote robustness.
  • Behavior adaptation: Thunderstorms can require slowing down, low-visibility and wet-surface countermeasures, and traction-mode shifts as road conditions change.The surveyed discussion links weather effects to demand for behavior adjustment.
  • Vehicle control: Crosswind produces side-slip, while wheel-encoder mismatch from slipping and traction loss threatens localization and successful control execution.Electronic stability control provides braking-based correction, but the discussion identifies delayed intervention as a concern.

9. Auxiliary Approaches in Adverse Conditions

Auxiliary approaches extend adverse-weather perception and decision-making beyond sensors mounted on the ego vehicle. The survey covers road-surface sensing, audio, roadside LiDAR, connected infrastructure, and external visibility aids.

  • Road surface detection: Road-surface detection can support weather classification because wetness and changing friction directly reflect weather-related road conditions.A vision-based DNN estimates road friction across dry, slippery, slurry, and icy surfaces.
  • Road surface detection: Audio-based road-surface detection learns from vehicle-speed, tire-surface, and wetness-dependent sounds using over 780,000 audio bins.The approach can use sounds from passing vehicles even when the ego vehicle is stationary.
  • Roadside units: Roadside LiDAR avoids vehicle-generated rain water screens and snow whirls, and processing with background filtering and object clustering improves detection in windy and snowy conditions.Connected systems can transmit roadside perception to AVs for planning.
  • Roadside units: Roadside LiDAR can provide redundant perception or weather classification in dense fog and other conditions where onboard sensors become unreliable.The surveyed systems quantify weather-specific point-cloud changes, but one study lacked fog data because of geographic conditions.
  • Connected infrastructure: V2X-enabled roadside units can provide nearby vehicles with weather classification and perception data from their own locations.The survey presents this as an advantage over relying on the AV alone.
  • Visibility aids: A smart automotive headlight recognized objects at 20 m with an 85% recognition rate while addressing visibility loss from rain or snow.The approach adapts a human-driver weather aid for autonomous-driving support.

10. Tools

Datasets, simulators, and controlled facilities provide essential coverage for adverse-weather autonomous-driving research. The survey highlights sparse real-world weather coverage, targeted dataset construction, safe simulation, and controllable test environments.

  • Datasets: Datasets are essential for extracting object-detection features and testing or validating autonomous-driving algorithms, but common training datasets contain few conditions beyond clear weather.Examples include rain in nuScenes, nighttime rain in A*3D, and strong light and shadow in ApolloScape.
  • Datasets: Weather datasets span rain, fog or haze, snow, strong light or night, and contamination, with sensor support varying across camera, thermal, radar, LiDAR, GNSS, and IMU platforms.The tabulated examples include CADCD, Oxford RobotCar, nuScenes, DENSE, WADS, and Boreas.
  • Datasets: Targeted collection can address regional gaps: CADC focuses on snow and combines eight cameras, LiDAR, and GNSS+INS in Canadian winter conditions.Its LiDAR was modified with a denoising method.
  • Simulators: CARLA enables custom complex road environments and virtually unlimited participants and scenarios that are difficult or costly to reproduce in field experiments.Simulation is particularly useful when seasonal or extreme weather is unavailable on demand.
  • Experimental facilities: Simulators provide zero-risk adverse-condition testing, while enclosed artificial tracks and weather chambers offer controllable precipitation rates with lower safety risks than ordinary field tests.The surveyed facilities include specialized proving grounds and cryospheric or weather simulators.

11. Trends, Limitations & Future Research

The survey identifies sensor fusion, machine learning, and V2X/IoT as central directions for improving adverse-weather autonomy, while highlighting coverage, hardware, and emerging-LiDAR limitations.

  • 11.2.1. Toward advanced sensor fusion: Sensor fusion combines complementary strengths of LiDAR, cameras, radar, infrared, gated, stereo, weather, and other sensors for weather-robust perception.The survey describes fusion as more reliable than relying on any single ADS sensor.
  • 11.2.1. Toward advanced sensor fusion: Future sensor configurations must balance performance against economic and computational cost, despite increasing onboard energy capacity.The survey recommends selecting feasible combinations of necessary sensors rather than adding every specialized sensor.
  • 11.2.2. Toward more sophisticated networks: Active learning can select unrecognized frames from unlabeled data for human annotation and subsequent training of adverse-weather perception networks.This process starts from a trained deep neural network and prioritizes difficult examples.
  • 11.2.3. Toward V2X and IoT: V2X extends perception beyond a vehicle’s direct line of sight, allowing earlier weather alerts, route changes, and safer decisions.Vehicles encountering weather or road-surface changes can relay assessments to vehicles approaching the location.
  • 11.2.3. Toward V2X and IoT: V2X-aided autonomous driving integrates vehicle, roadside, route, velocity, and traffic data, but remains limited by network and infrastructure coverage.The described system combines beyond-line-of-sight perception with extended planning.
  • 11.2.3. Toward V2X and IoT: V2X-d combines V2V and V2I to address limitations including V2V’s restricted horizontal view and roadside-camera vulnerability under weather.The architecture supports vehicle-density estimation in urban areas under all weather conditions.
  • 11.3. Limitations & Future Research: Sensor degradation from lifetime effects, temperature, and humidity remains insufficiently analyzed despite narrow operating tolerances.Mechanical changes can knock roof-, body-, windshield-, grille-, and bumper-mounted sensors out of calibration.

12. Conclusion

The conclusion summarizes the survey’s broad review of adverse-weather effects, mitigation methods, and supporting resources. It reports progress in wet-weather driving while identifying snow, strong light, contamination, and LiDAR robustness as continuing research needs.

  • 12. Conclusion: The survey reviews adverse-weather effects on major ADS sensors, hardware and mechanical solutions, perception enhancement, machine learning, image processing, and auxiliary systems.It also covers classification, assessment, control, planning, datasets, simulators, and experimental facilities.
  • 12. Conclusion: Rain and fog performance has improved, but LiDAR still needs improvement; snow remains focused on dataset expansion and perception enhancement.The authors identify extreme-snow point-cloud processing and interaction scenarios as future work.
  • 12. Conclusion: Strong light and contamination remain comparatively under-researched, while radar and thermal cameras reinforce robustness through sensor-fusion modalities.The conclusion frames fusion as support for adverse-weather ADS reliability.

Declaration of Interests

The authors report no known competing financial interests or personal relationships that could have influenced the work.

  • Declaration of Interests: The authors declare no known competing financial interests or personal relationships that could have influenced the reported work.

all-weather capabilities. URL: https://www.laserfocusworld. com/lasers-sources/article/14035383/frequencymodulated-

The referenced material concerns continuous-wave LiDAR’s claimed all-weather capabilities.

  • all-weather capabilities: The referenced material is identified by a URL concerning continuous-wave LiDAR and all-weather capabilities.
Loading 2112.08936v2…