Traditional sampling-based path planning algorithms,such as the rapidly-exploring random tree star(RRT^(*)),encounter critical limitations in unstructured orchard environments,including low sampling efficiency in narr...Traditional sampling-based path planning algorithms,such as the rapidly-exploring random tree star(RRT^(*)),encounter critical limitations in unstructured orchard environments,including low sampling efficiency in narrow passages,slow convergence,and high computational costs.To address these challenges,this paper proposes a novel hybrid global path planning algorithm integrating Gaussian sampling and quadtree optimization(RRT^(*)-GSQ).This methodology aims to enhance path planning by synergistically combining a Gaussian mixture sampling strategy to improve node generation in critical regions,an adaptive step-size and direction optimization mechanism for enhanced obstacle avoidance,a Quadtree-AABB collision detection framework to lower computational complexity,and a dynamic iteration control strategy for more efficient convergence.In obstacle-free and obstructed scenarios,compared with the conventional RRT^(*),the proposed algorithm reduced the number of node evaluations by 67.57%and 62.72%,and decreased the search time by 79.72%and 78.52%,respectively.In path tracking tests,the proposed algorithm achieved substantial reductions in RMSE of the final path compared to the conventional RRT^(*).Specifically,the lateral RMSE was reduced by 41.5%in obstacle-free environments and 59.3%in obstructed environments,while the longitudinal RMSE was reduced by 57.2%and 58.5%,respectively.Furthermore,the maximum absolute errors in both lateral and longitudinal directions were constrained within 0.75 m.Field validation experiments in an operational orchard confirmed the algorithm's practical effectiveness,showing reductions in the mean tracking error of 47.6%(obstacle-free)and 58.3%(with obstructed),alongside a 5.1%and 7.2%shortening of the path length compared to the baseline method.The proposed algorithm effectively enhances path planning efficiency and navigation accuracy for robots,presenting a superior solution for high-precision autonomous navigation of agricultural robots in orchard environments and holding significant value for engineering applications.展开更多
Dear Editor,This letter studies the motion planning issue for an autonomous underwater vehicle(AUV)in obstacle environment.We propose a novel integrated detection-communication waveform that enables simultaneous obsta...Dear Editor,This letter studies the motion planning issue for an autonomous underwater vehicle(AUV)in obstacle environment.We propose a novel integrated detection-communication waveform that enables simultaneous obstacle detection and self-localization.展开更多
Altun Mountains National Nature Reserve Established in 1983,the Altun Mountains National Nature Reserve is located in the eastern Kunlun Mountains,within Ruoqiang County of the Bayingolin Mongol Autonomous Prefecture ...Altun Mountains National Nature Reserve Established in 1983,the Altun Mountains National Nature Reserve is located in the eastern Kunlun Mountains,within Ruoqiang County of the Bayingolin Mongol Autonomous Prefecture in Xinjiang Uygur Autonomous Region.Covering a vast area of 45,000 square km,it stands as one of China’s largest and most pristine protected areas.With an average elevation of 4,580 metres,it represents a quintessential plateau desert ecosystem.展开更多
As 2026 unfolds,the image of an autonomous port in Ningbo loading a U.S.-bound ship encapsulates the transformative power of China’s surge in artificial intelligence(AI).This automation marvel not only symbolizes Ch...As 2026 unfolds,the image of an autonomous port in Ningbo loading a U.S.-bound ship encapsulates the transformative power of China’s surge in artificial intelligence(AI).This automation marvel not only symbolizes China’s economic power but also demonstrates how deeply integrated AI has become into its global competition strategy.Once largely confined to academic research and pilot programs,AI is now embedded in China’s national strategy.The goal:drive productivity,modernize industries,and strengthen its competitive position in global technology markets.展开更多
Diatoms,as natural sources of porous silica,have important potential for biomedical applications.Biohybrid microrobots also show promise for targeted delivery;however,research on converting diatoms into biohybrid micr...Diatoms,as natural sources of porous silica,have important potential for biomedical applications.Biohybrid microrobots also show promise for targeted delivery;however,research on converting diatoms into biohybrid microrobots and exploiting their intrinsic properties for cancer treatment remains limited.In this study,Thalassiosira weissflogii was transformed into biohybrid microrobots(Mag-Diatoms)while retaining its natural chlorophyll,thereby enabling Mag-Diatom-mediated photodynamic therapy(PDT)without additional drug modification.In this system,Mag-Diatoms act ed as microrobots,and their intrinsic chlorophyll serve d as a photosensitizer,exhibiting excellent biological safety.The autonomous closed-loop motion of the Mag-Diatoms was achieved using an artificial intelligence algorithm,which enabled controlled navigation along a preset trajectory.Mag-Diatoms also exhibited the ability to traverse narrow slits and target cancer cells within a cellular environment.The PDT effect was validated in vitro using human malignant glioblastoma(GBM)cell lines and primary cells derived from patients.The results revealed that the cell viability was closely related to the Mag-Diatom concentration,laser intensity,and irradiation time.Under combined Mag-Diatoms and laser treatment,viability decreased to 19.5%in primary cells and 3.6%in cell line models.Moreover,in vivo experiments using a mouse glioma model revealed that Mag-Diatom-mediated PDT effectively suppressed GBM progression.These findings highlight the potential of diatom-derived biohybrid microrobots,leveraging their natural properties,as a novel material and solution for PDT-based GBM therapy.展开更多
Unmanned aerial vehicles(UAVs),especially quadcopters,have become indispensable in numerous industrial and scientific applications due to their flexibility,lowcost,and capability to operate in dynamic environments.Thi...Unmanned aerial vehicles(UAVs),especially quadcopters,have become indispensable in numerous industrial and scientific applications due to their flexibility,lowcost,and capability to operate in dynamic environments.This paper presents a complete design and implementation of a compact autonomous quadcopter capable of trajectory tracking,object detection,precision landing,and real-time telemetry via long-range communication protocols.The system integrates an onboard flight controller running real-time sensor fusion algorithms,a vision-based detection system on a companion single-board computer,and a telemetry unit using Long Range(LoRa)communication.Extensive flight tests were conducted to validate the system’s stability,communication range,and autonomous capabilities.Potential applications in law enforcement,agriculture,search and rescue,and environmental monitoring are also discussed.展开更多
This paper proposes a deep learning-based 3D LiDAR perception framework designed for applications such as autonomous robots and vehicles.To address the high dependency on large-scale annotated data—an inherent limita...This paper proposes a deep learning-based 3D LiDAR perception framework designed for applications such as autonomous robots and vehicles.To address the high dependency on large-scale annotated data—an inherent limitation of deep learning models—this study introduces a hybrid perception architecture that incorporates expertdriven LiDAR processing techniques into the deep neural network.Traditional 3DLiDAR processingmethods typically remove ground planes and apply distance-or density-based clustering for object detection.In this work,such expert knowledge is encoded as feature-level inputs and fused with the deep network,therebymitigating the data dependency issue of conventional learning-based approaches.Specifically,the proposedmethod combines two expert algorithms—Patchwork++for ground segmentation and DBSCAN for clustering—with a PointPillars-based LiDAR detection network.We design four hybrid versions of the network depending on the stage and method of integrating expert features into the feature map of the deep model.Among these,Version 4 incorporates a modified neck structure in PointPillars and introduces a new Cluster 2D Pseudo-Map Branch that utilizes cluster-level pseudo-images generated from Patchwork++and DBSCAN.This version achieved a+3.88%improvement mean Average Precision(mAP)compared to the baseline PointPillars.The results demonstrate that embedding expert-based perception logic into deep neural architectures can effectively enhance performance and reduce dependency on extensive training datasets,offering a promising direction for robust 3D LiDAR object detection in real-world scenarios.展开更多
Autonomous vehicles rely heavily on accurate and efficient scene segmentation for safe navigation and efficient operations.Traditional Bird’s Eye View(BEV)methods on semantic scene segmentation,which leverage multimo...Autonomous vehicles rely heavily on accurate and efficient scene segmentation for safe navigation and efficient operations.Traditional Bird’s Eye View(BEV)methods on semantic scene segmentation,which leverage multimodal sensor fusion,often struggle with noisy data and demand high-performance GPUs,leading to sensor misalignment and performance degradation.This paper introduces an Enhanced Channel Attention BEV(ECABEV),a novel approach designed to address the challenges under insufficient GPU memory conditions.ECABEV integrates camera and radar data through a de-noise enhanced channel attention mechanism,which utilizes global average and max pooling to effectively filter out noise while preserving discriminative features.Furthermore,an improved fusion approach is proposed to efficiently merge categorical data across modalities.To reduce computational overhead,a bilinear interpolation layer normalizationmethod is devised to ensure spatial feature fidelity.Moreover,a scalable crossentropy loss function is further designed to handle the imbalanced classes with less computational efficiency sacrifice.Extensive experiments on the nuScenes dataset demonstrate that ECABEV achieves state-of-the-art performance with an IoU of 39.961,using a lightweight ViT-B/14 backbone and lower resolution(224×224).Our approach highlights its cost-effectiveness and practical applicability,even on low-end devices.The code is publicly available at:https://github.com/YYF-CQU/ECABEV.git.展开更多
Scalable simulation leveraging real-world data plays an essential role in advancing autonomous driving,owing to its efficiency and applicability in both training and evaluating algorithms.Consequently,there has been i...Scalable simulation leveraging real-world data plays an essential role in advancing autonomous driving,owing to its efficiency and applicability in both training and evaluating algorithms.Consequently,there has been increasing attention on generating highly realistic and consistent driving videos,particularly those involving viewpoint changes guided by the control commands or trajectories of ego vehicles.However,current reconstruction approaches,such as Neural Radiance Fields and 3D Gaussian Splatting,frequently suffer from limited generalization and depend on substantial input data.Meanwhile,2D generative models,though capable of producing unknown scenes,still have room for improvement in terms of coherence and visual realism.To overcome these challenges,we introduce GenScene,a world model that synthesizes front-view driving videos conditioned on trajectories.A new temporal module is presented to improve video consistency by extracting the global context of each frame,calculating relationships of frames using these global representations,and fusing frame contexts accordingly.Moreover,we propose an innovative attention mechanism that computes relations of pixels within each frame and pixels in the corresponding window range of the initial frame.Extensive experiments show that our approach surpasses various state-of-the-art models in driving video generation,and the introduced modules contribute significantly to model performance.This work establishes a new paradigm for goal-oriented video synthesis in autonomous driving,which facilitates on-demand simulation to expedite algorithm development.展开更多
Autonomous connected vehicles(ACV)involve advanced control strategies to effectively balance safety,efficiency,energy consumption,and passenger comfort.This research introduces a deep reinforcement learning(DRL)-based...Autonomous connected vehicles(ACV)involve advanced control strategies to effectively balance safety,efficiency,energy consumption,and passenger comfort.This research introduces a deep reinforcement learning(DRL)-based car-following(CF)framework employing the Deep Deterministic Policy Gradient(DDPG)algorithm,which integrates a multi-objective reward function that balances the four goals while maintaining safe policy learning.Utilizing real-world driving data from the highD dataset,the proposed model learns adaptive speed control policies suitable for dynamic traffic scenarios.The performance of the DRL-based model is evaluated against a traditional model predictive control-adaptive cruise control(MPC-ACC)controller.Results show that theDRLmodel significantly enhances safety,achieving zero collisions and a higher average time-to-collision(TTC)of 8.45 s,compared to 5.67 s for MPC and 6.12 s for human drivers.For efficiency,the model demonstrates 89.2% headway compliance and maintains speed tracking errors below 1.2 m/s in 90% of cases.In terms of energy optimization,the proposed approach reduces fuel consumption by 5.4% relative to MPC.Additionally,it enhances passenger comfort by lowering jerk values by 65%,achieving 0.12 m/s3 vs.0.34 m/s3 for human drivers.A multi-objective reward function is integrated to ensure stable policy convergence while simultaneously balancing the four key performance metrics.Moreover,the findings underscore the potential of DRL in advancing autonomous vehicle control,offering a robust and sustainable solution for safer,more efficient,and more comfortable transportation systems.展开更多
Embodied intelligence is redefining policing On the first day of 2026 chunyun,a period of high mobility associated with the Chinese New Year,the city of Jingzhou in Hubei Province welcomed new participants in road saf...Embodied intelligence is redefining policing On the first day of 2026 chunyun,a period of high mobility associated with the Chinese New Year,the city of Jingzhou in Hubei Province welcomed new participants in road safety:police robots capable of moving autonomously and interacting with passengers.Deployed on a trial basis on 2 February,these robots quickly demonstrated their usefulness in various urban settings.展开更多
This article studies the problem of image segmentation-based semantic communication in autonomous driving.In real traffic scenes,the detecting of objects(e.g.,vehicles and pedestrians)is more important to guarantee dr...This article studies the problem of image segmentation-based semantic communication in autonomous driving.In real traffic scenes,the detecting of objects(e.g.,vehicles and pedestrians)is more important to guarantee driving safety,which is always ignored in existing works.Therefore,we propose a vehicular image segmentation-oriented semantic communication system,termed VIS-SemCom,focusing on transmitting and recovering image semantic features of high-important objects to reduce transmission redundancy.First,we develop a semantic codec based on Swin Transformer architecture,which expands the perceptual field thus improving the segmentation accuracy.To highlight the important objects'accuracy,we propose a multi-scale semantic extraction method by assigning the number of Swin Transformer blocks for diverse resolution semantic features.Also,an importance-aware loss incorporating important levels is devised,and an online hard example mining(OHEM)strategy is proposed to handle small sample issues in the dataset.Finally,experimental results demonstrate that the proposed VIS-SemCom can achieve a significant mean intersection over union(mIoU)performance in the SNR regions,a reduction of transmitted data volume by about 60%at 60%mIoU,and improve the segmentation accuracy of important objects,compared to baseline image communication.展开更多
Addressing optimal confrontation methods in multi-agent attack-defense scenarios is a complex challenge.Multi-Agent Reinforcement Learning(MARL)provides an effective framework for tackling sequential decision-making p...Addressing optimal confrontation methods in multi-agent attack-defense scenarios is a complex challenge.Multi-Agent Reinforcement Learning(MARL)provides an effective framework for tackling sequential decision-making problems,significantly enhancing swarm intelligence in maneuvering.However,applying MARL to unmanned swarms presents two primary challenges.First,defensive agents must balance autonomy with collaboration under limited perception while coordinating against adversaries.Second,current algorithms aim to maximize global or individual rewards,making them sensitive to fluctuations in enemy strategies and environmental changes,especially when rewards are sparse.To tackle these issues,we propose an algorithm of MultiAgent Reinforcement Learning with Layered Autonomy and Collaboration(MARL-LAC)for collaborative confrontations.This algorithm integrates dual twin Critics to mitigate the high variance associated with policy gradients.Furthermore,MARL-LAC employs layered autonomy and collaboration to address multi-objective problems,specifically learning a global reward function for the swarm alongside local reward functions for individual defensive agents.Experimental results demonstrate that MARL-LAC enhances decision-making and collaborative behaviors among agents,outperforming the existing algorithms and emphasizing the importance of layered autonomy and collaboration in multi-agent systems.The observed adversarial behaviors demonstrate that agents using MARL-LAC effectively maintain cohesive formations that conceal their intentions by confusing the offensive agent while successfully encircling the target.展开更多
Autonomous vehicles operate without direct human intervention,which introduces safety risks that differ from those of conventional vehicles.Although many studies have examined safety issues related to autonomous drivi...Autonomous vehicles operate without direct human intervention,which introduces safety risks that differ from those of conventional vehicles.Although many studies have examined safety issues related to autonomous driving,high-risk situations have often been defined using single indicators,making it difficult to capture the complex and evolving nature of accident risk.To address this limitation,this study proposes a structured framework for defining and analyzing high-risk situations throughout the traffic accident process.High-risk situations are described using three complementary indicators:accident likelihood,accident severity,and accident duration.These indicators explain how risk emerges,increases,and persists over time.Based on this concept,a framework for traffic accident visualization analysis is developed to support phase-specific risk assessment and visualization.The framework combines accident-phase information with factor-level risk contributions,allowing systematic identification of key factors and their interactions across different accident stages.Using combinations of the three indicators,high-risk situations are classified into twenty-seven distinct types,providing a clear typology for complex accident scenarios involving autonomous vehicles.The applicability of the proposed framework is demonstrated through two representative accident scenarioswith different risk characteristics.The results showthat the framework effectively captures interactions among multiple risk factors,explains how risk levels change from pre-crash to post-crash phases,and identifies contributing factors that are difficult to detect using conventional traffic accident investigation methods.Overall,the proposed framework offers a practical basis for autonomous vehicle accident analysis,safety evaluation,and policy-related decision-making.展开更多
In intelligent transportation systems,object detection for a surveillance video is one of the important functions.The performance of existing surveillance video object detection algorithms is affected by the conflict ...In intelligent transportation systems,object detection for a surveillance video is one of the important functions.The performance of existing surveillance video object detection algorithms is affected by the conflict between the features of the objects,which leads to a decline in precision.Therefore,an object detection algorithm based on deep learning and salient feature fusion is proposed.The proposed method introduces a non-weight-sharing network to process the salient features of the image and fuse them with the features extracted from the red blue green branch.Different from the previous solutions,the salient feature extraction branch uses the boundary features and statistical features of the image and fuses the features of the two branches in the efficient layer aggregation networks structure.At the same time,the attention module is used in efficient layer aggregation networks with convolutional block attention module to improve the efficiency of feature utilisation.The training and evaluation are carried out in the constructed surveillance video feature conflict dataset,and eight scenes are constructed in the way of orthogonal experiments.The experimental results show that the performance of object detection can be significantly improved by using the proposed method in the object detection task of the intelligent transportation system surveillance video feature conflict scene.展开更多
To systematically elucidate the central role of the cardiac autonomic nervous system(ANS)in maintaining cardiovascular homeostasis,analyze the pathological mechanisms underlying its dysregulation,integrate multidiscip...To systematically elucidate the central role of the cardiac autonomic nervous system(ANS)in maintaining cardiovascular homeostasis,analyze the pathological mechanisms underlying its dysregulation,integrate multidisciplinary research findings from the U.S.stimulating peripheral nerves to alleviate disease symptoms(SPARC)program,and evaluate the unique advantages and current limitations of acupuncture in modulating cardiac ANS function,ultimately proposing novel strategies for the prevention and treatment of cardiovascular diseases.A systematic literature review was conducted to synthesize current knowledge on the fundamental regulatory mechanisms of the cardiac ANS,the SPARC program's innovative contributions to neuroanatomy and neural pathway mapping,as well as clinical and experimental evidence supporting acupuncture's modulation of the ANS.The SPARC program has made significant progress in elucidating the anatomical organization and neural circuitry of the cardiac ANS through interdisciplinary collaboration,offering novel insights and methodological frameworks for studying cardiac autonomic regulation.Acupuncture,as a cornerstone of traditional medicine,has demonstrated both specificity and multi-target regulatory effects on the cardiac ANS through clinical and experimental studies.However,challenges remain,including an incomplete mechanistic understanding,technical limitations in research methodologies,and difficulties in translating findings into clinical practice.Future research on acupuncture should build upon its inherent strengths while advancing its integration with modern scientific and technological approaches.Strengthening interdisciplinary collaboration and leveraging artificial intelligence can open new frontiers in mechanistic exploration and technological innovation.These efforts will facilitate the internationalization of acupuncture research and contribute innovative perspectives and therapeutic strategies for cardiovascular disease prevention and treatment.展开更多
As urban landscapes evolve and vehicular volumes soar,traditional traffic monitoring systems struggle to scale,often failing under the complexities of dense,dynamic,and occluded environments.This paper introduces a no...As urban landscapes evolve and vehicular volumes soar,traditional traffic monitoring systems struggle to scale,often failing under the complexities of dense,dynamic,and occluded environments.This paper introduces a novel,unified deep learning framework for vehicle detection,tracking,counting,and classification in aerial imagery designed explicitly for modern smart city infrastructure demands.Our approach begins with adaptive histogram equalization to optimize aerial image clarity,followed by a cutting-edge scene parsing technique using Mask2Former,enabling robust segmentation even in visually congested settings.Vehicle detection leverages the latest YOLOv11 architecture,delivering superior accuracy in aerial contexts by addressing occlusion,scale variance,and fine-grained object differentiation.We incorporate the highly efficient ByteTrack algorithm for tracking,enabling seamless identity preservation across frames.Vehicle counting is achieved through an unsupervised DBSCAN-based method,ensuring adaptability to varying traffic densities.We further introduce a hybrid feature extraction module combining Convolutional Neural Networks(CNNs)with Zernike Moments,capturing both deep semantic and geometric signatures of vehicles.The final classification is powered by NASNet,a neural architecture search-optimized model,ensuring high accuracy across diverse vehicle types and orientations.Extensive evaluations of the VAID benchmark dataset demonstrate the system’s outstanding performance,achieving 96%detection,94%tracking,and 96.4%classification accuracy.On the UAVDT dataset,the system attains 95%detection,93%tracking,and 95%classification accuracy,confirming its robustness across diverse aerial traffic scenarios.These results establish new benchmarks in aerial traffic analysis and validate the framework’s scalability,making it a powerful and adaptable solution for next-generation intelligent transportation systems and urban surveillance.展开更多
This survey presents a comprehensive examination of sensor fusion research spanning four decades,tracing the methodological evolution,application domains,and alignment with classical hierarchical models.Building on th...This survey presents a comprehensive examination of sensor fusion research spanning four decades,tracing the methodological evolution,application domains,and alignment with classical hierarchical models.Building on this long-term trajectory,the foundational approaches such as probabilistic inference,early neural networks,rulebasedmethods,and feature-level fusion established the principles of uncertainty handling andmulti-sensor integration in the 1990s.The fusion methods of 2000s marked the consolidation of these ideas through advanced Kalman and particle filtering,Bayesian–Dempster–Shafer hybrids,distributed consensus algorithms,and machine learning ensembles for more robust and domain-specific implementations.From 2011 to 2020,the widespread adoption of deep learning transformed the field driving some major breakthroughs in the autonomous vehicles domain.A key contribution of this work is the assessment of contemporary methods against the JDL model,revealing gaps at higher levels-especially in situation and impact assessment.Contemporary methods offer only limited implementation of higher-level fusion.The survey also reviews the benchmark multi-sensor datasets,noting their role in advancing the field while identifying major shortcomings like the lack of domain diversity and hierarchical coverage.By synthesizing developments across decades and paradigms,this survey provides both a historical narrative and a forward-looking perspective.It highlights unresolved challenges in transparency,scalability,robustness,and trustworthiness,while identifying emerging paradigms such as neuromorphic fusion and explainable AI as promising directions.This paves the way forward for advancing sensor fusion towards transparent and adaptive next-generation autonomous systems.展开更多
Analyzing the driving behavior of autonomous vehicles(AV)in mixed traffic conditions at urban intersections has become increasingly important for improving intersection design,providing infrastructure-based guidance i...Analyzing the driving behavior of autonomous vehicles(AV)in mixed traffic conditions at urban intersections has become increasingly important for improving intersection design,providing infrastructure-based guidance information,and developing capability-enhanced AV perception systems.This study investigated the contributing factors affecting AV driving behavior using theWaymo Open Dataset.Binarized autonomous driving stability metrics,derived via a kernel density estimation,served as the target variables for a random forest classification model.The model’s input variables included 15 factors divided into four types:intersection-related,surrounding object-related,road infrastructure-related,and time-of-day-related types.The random forest classification model was employed to identify the key factors affecting autonomous driving behavior.In addition,the identified factors were further ranked based on feature importance.SHAP analysis was utilized to enhance model interpretability by quantifying the contribution of each factor and identifying their directional impacts.The type of intersection factor was found to have an importance of 0.243 and was the most influential factor on autonomous driving behavior.On average,intersection-related factors had an importance of 0.196,which is approximately a 31.1%margin over the average importance of surrounding object-related factors.Additionally,the surrounding object-related factors that were collected through sensors on the autonomous vehicle had a high degree of feature importance,especially with the number of pedestrians having the highest importance(0.107)of the types of objects.The correlation between these findings can contribute to the development of various treatments to improvemore harmonized AVs’maneuvering with other road users and facilities in urban mixed traffic environments.展开更多
This study aimed to enhance the performance of semantic segmentation for autonomous driving by improving the 2DPASS model.Two novel improvements were proposed and implemented in this paper:dynamically adjusting the lo...This study aimed to enhance the performance of semantic segmentation for autonomous driving by improving the 2DPASS model.Two novel improvements were proposed and implemented in this paper:dynamically adjusting the loss function ratio and integrating an attention mechanism(CBAM).First,the loss function weights were adjusted dynamically.The grid search method is used for deciding the best ratio of 7:3.It gives greater emphasis to the cross-entropy loss,which resulted in better segmentation performance.Second,CBAM was applied at different layers of the 2Dencoder.Heatmap analysis revealed that introducing it after the second block of 2D image encoding produced the most effective enhancement of important feature representation.The training epoch was chosen for optimizing the best value by experiments,which improved model convergence and overall accuracy.To evaluate the proposed approach,experiments were conducted based on the SemanticKITTI database.The results showed that the improved model achieved higher segmentation accuracy by 64.31%,improved 11.47% in mIoU compared with the conventional 2DPASS model(baseline:52.84%).It was more effective at detecting small and distant objects and clearly identifying boundaries between different classes.Issues such as noise and variations in data distribution affected its accuracy,indicating the need for further refinement.Overall,the proposed improvements to the 2DPASS model demonstrated the potential to advance semantic segmentation technology and contributed to a more reliable perception of complex,dynamic environments in autonomous vehicles.Accurate segmentation enhances the vehicle’s ability to distinguish different objects,and this improvement directly supports safer navigation,robust decision-making,and efficient path planning,making it highly applicable to real-world deployment of autonomous systems in urban and highway settings.展开更多
基金National Natural Science Foundation of China(32301712)Natural Science Foundation of Jiangsu Province(BK20230548,BK20250876)+2 种基金Project of Faculty of Agricultural Equipment of Jiangsu University(NGXB20240203)A Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions(PAPD-2023-87)Open Funding Project of the Key Laboratory of Modern Agricultural Equipment and Technology(Jiangsu University),Ministry of Education(MAET202101)。
文摘Traditional sampling-based path planning algorithms,such as the rapidly-exploring random tree star(RRT^(*)),encounter critical limitations in unstructured orchard environments,including low sampling efficiency in narrow passages,slow convergence,and high computational costs.To address these challenges,this paper proposes a novel hybrid global path planning algorithm integrating Gaussian sampling and quadtree optimization(RRT^(*)-GSQ).This methodology aims to enhance path planning by synergistically combining a Gaussian mixture sampling strategy to improve node generation in critical regions,an adaptive step-size and direction optimization mechanism for enhanced obstacle avoidance,a Quadtree-AABB collision detection framework to lower computational complexity,and a dynamic iteration control strategy for more efficient convergence.In obstacle-free and obstructed scenarios,compared with the conventional RRT^(*),the proposed algorithm reduced the number of node evaluations by 67.57%and 62.72%,and decreased the search time by 79.72%and 78.52%,respectively.In path tracking tests,the proposed algorithm achieved substantial reductions in RMSE of the final path compared to the conventional RRT^(*).Specifically,the lateral RMSE was reduced by 41.5%in obstacle-free environments and 59.3%in obstructed environments,while the longitudinal RMSE was reduced by 57.2%and 58.5%,respectively.Furthermore,the maximum absolute errors in both lateral and longitudinal directions were constrained within 0.75 m.Field validation experiments in an operational orchard confirmed the algorithm's practical effectiveness,showing reductions in the mean tracking error of 47.6%(obstacle-free)and 58.3%(with obstructed),alongside a 5.1%and 7.2%shortening of the path length compared to the baseline method.The proposed algorithm effectively enhances path planning efficiency and navigation accuracy for robots,presenting a superior solution for high-precision autonomous navigation of agricultural robots in orchard environments and holding significant value for engineering applications.
基金supported in part by the National Natural Science Foundation of China(U25A20473,62222314)the YanZhao Young Scientist Project of Hebei Province(F2024203047)+2 种基金the Natural Science Foundation of Hebei Province(F2022203001,F2024203072)the State Key Laboratory of Submarine Geoscience(sglkt2025-7)the Education Department Foundation of Hebei Province(JCZX2025027)。
文摘Dear Editor,This letter studies the motion planning issue for an autonomous underwater vehicle(AUV)in obstacle environment.We propose a novel integrated detection-communication waveform that enables simultaneous obstacle detection and self-localization.
文摘Altun Mountains National Nature Reserve Established in 1983,the Altun Mountains National Nature Reserve is located in the eastern Kunlun Mountains,within Ruoqiang County of the Bayingolin Mongol Autonomous Prefecture in Xinjiang Uygur Autonomous Region.Covering a vast area of 45,000 square km,it stands as one of China’s largest and most pristine protected areas.With an average elevation of 4,580 metres,it represents a quintessential plateau desert ecosystem.
文摘As 2026 unfolds,the image of an autonomous port in Ningbo loading a U.S.-bound ship encapsulates the transformative power of China’s surge in artificial intelligence(AI).This automation marvel not only symbolizes China’s economic power but also demonstrates how deeply integrated AI has become into its global competition strategy.Once largely confined to academic research and pilot programs,AI is now embedded in China’s national strategy.The goal:drive productivity,modernize industries,and strengthen its competitive position in global technology markets.
基金supported by the National Key R&D Program of China(No.2023YFB4705600)the National Natural Science Foundation of China(Nos.U23A20342,U20A20380,62273331,62127811,and 82373342)+4 种基金CAS Project for Young Scientists in Basic Research(No.YSBR-036)New Cornerstone Science Foundation through the XPLORER PRIZE,CAS/SAFEA International Partnership Program for Creative Research Teams,the Science and Technology Planning Project of Liaoning Province(No.2021JH1/10400049)Shengjing Hospital of China Medical University 345 Talent Project(No.1000801592)the Joint Project of Liaoning Province(No.2023JH2/101700202)“the Fundamental Research Funds for the Central Universities”,South-Central Minzu University(No.CZQ 25014).
文摘Diatoms,as natural sources of porous silica,have important potential for biomedical applications.Biohybrid microrobots also show promise for targeted delivery;however,research on converting diatoms into biohybrid microrobots and exploiting their intrinsic properties for cancer treatment remains limited.In this study,Thalassiosira weissflogii was transformed into biohybrid microrobots(Mag-Diatoms)while retaining its natural chlorophyll,thereby enabling Mag-Diatom-mediated photodynamic therapy(PDT)without additional drug modification.In this system,Mag-Diatoms act ed as microrobots,and their intrinsic chlorophyll serve d as a photosensitizer,exhibiting excellent biological safety.The autonomous closed-loop motion of the Mag-Diatoms was achieved using an artificial intelligence algorithm,which enabled controlled navigation along a preset trajectory.Mag-Diatoms also exhibited the ability to traverse narrow slits and target cancer cells within a cellular environment.The PDT effect was validated in vitro using human malignant glioblastoma(GBM)cell lines and primary cells derived from patients.The results revealed that the cell viability was closely related to the Mag-Diatom concentration,laser intensity,and irradiation time.Under combined Mag-Diatoms and laser treatment,viability decreased to 19.5%in primary cells and 3.6%in cell line models.Moreover,in vivo experiments using a mouse glioma model revealed that Mag-Diatom-mediated PDT effectively suppressed GBM progression.These findings highlight the potential of diatom-derived biohybrid microrobots,leveraging their natural properties,as a novel material and solution for PDT-based GBM therapy.
文摘Unmanned aerial vehicles(UAVs),especially quadcopters,have become indispensable in numerous industrial and scientific applications due to their flexibility,lowcost,and capability to operate in dynamic environments.This paper presents a complete design and implementation of a compact autonomous quadcopter capable of trajectory tracking,object detection,precision landing,and real-time telemetry via long-range communication protocols.The system integrates an onboard flight controller running real-time sensor fusion algorithms,a vision-based detection system on a companion single-board computer,and a telemetry unit using Long Range(LoRa)communication.Extensive flight tests were conducted to validate the system’s stability,communication range,and autonomous capabilities.Potential applications in law enforcement,agriculture,search and rescue,and environmental monitoring are also discussed.
基金supported by Basic Science Research Program through the National Research Foundation of Korea(NRF)funded by the Ministry of Education(RS-2023-00245084)by Korea Institute for Advancement of Technology(KIAT)grant funded by the Korea Government(MOTIE)(RS-2024-00415938,HRD Program for Industrial Innovation)and Soonchunhyang University.
文摘This paper proposes a deep learning-based 3D LiDAR perception framework designed for applications such as autonomous robots and vehicles.To address the high dependency on large-scale annotated data—an inherent limitation of deep learning models—this study introduces a hybrid perception architecture that incorporates expertdriven LiDAR processing techniques into the deep neural network.Traditional 3DLiDAR processingmethods typically remove ground planes and apply distance-or density-based clustering for object detection.In this work,such expert knowledge is encoded as feature-level inputs and fused with the deep network,therebymitigating the data dependency issue of conventional learning-based approaches.Specifically,the proposedmethod combines two expert algorithms—Patchwork++for ground segmentation and DBSCAN for clustering—with a PointPillars-based LiDAR detection network.We design four hybrid versions of the network depending on the stage and method of integrating expert features into the feature map of the deep model.Among these,Version 4 incorporates a modified neck structure in PointPillars and introduces a new Cluster 2D Pseudo-Map Branch that utilizes cluster-level pseudo-images generated from Patchwork++and DBSCAN.This version achieved a+3.88%improvement mean Average Precision(mAP)compared to the baseline PointPillars.The results demonstrate that embedding expert-based perception logic into deep neural architectures can effectively enhance performance and reduce dependency on extensive training datasets,offering a promising direction for robust 3D LiDAR object detection in real-world scenarios.
基金funded by the National Natural Science Foundation of China,grant number 62262045the Fundamental Research Funds for the Central Universities,grant number 2023CDJYGRH-YB11the Open Funding of SUGON Industrial Control and Security Center,grant number CUIT-SICSC-2025-03.
文摘Autonomous vehicles rely heavily on accurate and efficient scene segmentation for safe navigation and efficient operations.Traditional Bird’s Eye View(BEV)methods on semantic scene segmentation,which leverage multimodal sensor fusion,often struggle with noisy data and demand high-performance GPUs,leading to sensor misalignment and performance degradation.This paper introduces an Enhanced Channel Attention BEV(ECABEV),a novel approach designed to address the challenges under insufficient GPU memory conditions.ECABEV integrates camera and radar data through a de-noise enhanced channel attention mechanism,which utilizes global average and max pooling to effectively filter out noise while preserving discriminative features.Furthermore,an improved fusion approach is proposed to efficiently merge categorical data across modalities.To reduce computational overhead,a bilinear interpolation layer normalizationmethod is devised to ensure spatial feature fidelity.Moreover,a scalable crossentropy loss function is further designed to handle the imbalanced classes with less computational efficiency sacrifice.Extensive experiments on the nuScenes dataset demonstrate that ECABEV achieves state-of-the-art performance with an IoU of 39.961,using a lightweight ViT-B/14 backbone and lower resolution(224×224).Our approach highlights its cost-effectiveness and practical applicability,even on low-end devices.The code is publicly available at:https://github.com/YYF-CQU/ECABEV.git.
基金supported by the Cultivation Program for Major Scientific Research Projects of Harbin Institute of Technology(ZDXMPY20180109).
文摘Scalable simulation leveraging real-world data plays an essential role in advancing autonomous driving,owing to its efficiency and applicability in both training and evaluating algorithms.Consequently,there has been increasing attention on generating highly realistic and consistent driving videos,particularly those involving viewpoint changes guided by the control commands or trajectories of ego vehicles.However,current reconstruction approaches,such as Neural Radiance Fields and 3D Gaussian Splatting,frequently suffer from limited generalization and depend on substantial input data.Meanwhile,2D generative models,though capable of producing unknown scenes,still have room for improvement in terms of coherence and visual realism.To overcome these challenges,we introduce GenScene,a world model that synthesizes front-view driving videos conditioned on trajectories.A new temporal module is presented to improve video consistency by extracting the global context of each frame,calculating relationships of frames using these global representations,and fusing frame contexts accordingly.Moreover,we propose an innovative attention mechanism that computes relations of pixels within each frame and pixels in the corresponding window range of the initial frame.Extensive experiments show that our approach surpasses various state-of-the-art models in driving video generation,and the introduced modules contribute significantly to model performance.This work establishes a new paradigm for goal-oriented video synthesis in autonomous driving,which facilitates on-demand simulation to expedite algorithm development.
基金the Hebei Province Science and Technology Plan Project(19221909D)rincess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R308),Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘Autonomous connected vehicles(ACV)involve advanced control strategies to effectively balance safety,efficiency,energy consumption,and passenger comfort.This research introduces a deep reinforcement learning(DRL)-based car-following(CF)framework employing the Deep Deterministic Policy Gradient(DDPG)algorithm,which integrates a multi-objective reward function that balances the four goals while maintaining safe policy learning.Utilizing real-world driving data from the highD dataset,the proposed model learns adaptive speed control policies suitable for dynamic traffic scenarios.The performance of the DRL-based model is evaluated against a traditional model predictive control-adaptive cruise control(MPC-ACC)controller.Results show that theDRLmodel significantly enhances safety,achieving zero collisions and a higher average time-to-collision(TTC)of 8.45 s,compared to 5.67 s for MPC and 6.12 s for human drivers.For efficiency,the model demonstrates 89.2% headway compliance and maintains speed tracking errors below 1.2 m/s in 90% of cases.In terms of energy optimization,the proposed approach reduces fuel consumption by 5.4% relative to MPC.Additionally,it enhances passenger comfort by lowering jerk values by 65%,achieving 0.12 m/s3 vs.0.34 m/s3 for human drivers.A multi-objective reward function is integrated to ensure stable policy convergence while simultaneously balancing the four key performance metrics.Moreover,the findings underscore the potential of DRL in advancing autonomous vehicle control,offering a robust and sustainable solution for safer,more efficient,and more comfortable transportation systems.
文摘Embodied intelligence is redefining policing On the first day of 2026 chunyun,a period of high mobility associated with the Chinese New Year,the city of Jingzhou in Hubei Province welcomed new participants in road safety:police robots capable of moving autonomously and interacting with passengers.Deployed on a trial basis on 2 February,these robots quickly demonstrated their usefulness in various urban settings.
基金National Natural Science Foundation of China under Grants No.62171047,U22B2001,62271065,62001051Beijing Natural Science Foundation under Grant L223027BUPT Excellent Ph.D Students Foundation under Grants CX2021114。
文摘This article studies the problem of image segmentation-based semantic communication in autonomous driving.In real traffic scenes,the detecting of objects(e.g.,vehicles and pedestrians)is more important to guarantee driving safety,which is always ignored in existing works.Therefore,we propose a vehicular image segmentation-oriented semantic communication system,termed VIS-SemCom,focusing on transmitting and recovering image semantic features of high-important objects to reduce transmission redundancy.First,we develop a semantic codec based on Swin Transformer architecture,which expands the perceptual field thus improving the segmentation accuracy.To highlight the important objects'accuracy,we propose a multi-scale semantic extraction method by assigning the number of Swin Transformer blocks for diverse resolution semantic features.Also,an importance-aware loss incorporating important levels is devised,and an online hard example mining(OHEM)strategy is proposed to handle small sample issues in the dataset.Finally,experimental results demonstrate that the proposed VIS-SemCom can achieve a significant mean intersection over union(mIoU)performance in the SNR regions,a reduction of transmitted data volume by about 60%at 60%mIoU,and improve the segmentation accuracy of important objects,compared to baseline image communication.
基金co-supported by the National Natural Science Foundation of China(Nos.72371052 and 71871042).
文摘Addressing optimal confrontation methods in multi-agent attack-defense scenarios is a complex challenge.Multi-Agent Reinforcement Learning(MARL)provides an effective framework for tackling sequential decision-making problems,significantly enhancing swarm intelligence in maneuvering.However,applying MARL to unmanned swarms presents two primary challenges.First,defensive agents must balance autonomy with collaboration under limited perception while coordinating against adversaries.Second,current algorithms aim to maximize global or individual rewards,making them sensitive to fluctuations in enemy strategies and environmental changes,especially when rewards are sparse.To tackle these issues,we propose an algorithm of MultiAgent Reinforcement Learning with Layered Autonomy and Collaboration(MARL-LAC)for collaborative confrontations.This algorithm integrates dual twin Critics to mitigate the high variance associated with policy gradients.Furthermore,MARL-LAC employs layered autonomy and collaboration to address multi-objective problems,specifically learning a global reward function for the swarm alongside local reward functions for individual defensive agents.Experimental results demonstrate that MARL-LAC enhances decision-making and collaborative behaviors among agents,outperforming the existing algorithms and emphasizing the importance of layered autonomy and collaboration in multi-agent systems.The observed adversarial behaviors demonstrate that agents using MARL-LAC effectively maintain cohesive formations that conceal their intentions by confusing the offensive agent while successfully encircling the target.
基金supported by the Korea Institute of Police Technology(No.:RS-2024-00405603).
文摘Autonomous vehicles operate without direct human intervention,which introduces safety risks that differ from those of conventional vehicles.Although many studies have examined safety issues related to autonomous driving,high-risk situations have often been defined using single indicators,making it difficult to capture the complex and evolving nature of accident risk.To address this limitation,this study proposes a structured framework for defining and analyzing high-risk situations throughout the traffic accident process.High-risk situations are described using three complementary indicators:accident likelihood,accident severity,and accident duration.These indicators explain how risk emerges,increases,and persists over time.Based on this concept,a framework for traffic accident visualization analysis is developed to support phase-specific risk assessment and visualization.The framework combines accident-phase information with factor-level risk contributions,allowing systematic identification of key factors and their interactions across different accident stages.Using combinations of the three indicators,high-risk situations are classified into twenty-seven distinct types,providing a clear typology for complex accident scenarios involving autonomous vehicles.The applicability of the proposed framework is demonstrated through two representative accident scenarioswith different risk characteristics.The results showthat the framework effectively captures interactions among multiple risk factors,explains how risk levels change from pre-crash to post-crash phases,and identifies contributing factors that are difficult to detect using conventional traffic accident investigation methods.Overall,the proposed framework offers a practical basis for autonomous vehicle accident analysis,safety evaluation,and policy-related decision-making.
基金supported by the National Key Research and Development Programme of China(2021YFB3202200)the National Natural Science Foundation of China(52072333)Hebei Provincial Department of Education in the postgraduate innovation ability training funding project(CXZZBS2023061).
文摘In intelligent transportation systems,object detection for a surveillance video is one of the important functions.The performance of existing surveillance video object detection algorithms is affected by the conflict between the features of the objects,which leads to a decline in precision.Therefore,an object detection algorithm based on deep learning and salient feature fusion is proposed.The proposed method introduces a non-weight-sharing network to process the salient features of the image and fuse them with the features extracted from the red blue green branch.Different from the previous solutions,the salient feature extraction branch uses the boundary features and statistical features of the image and fuses the features of the two branches in the efficient layer aggregation networks structure.At the same time,the attention module is used in efficient layer aggregation networks with convolutional block attention module to improve the efficiency of feature utilisation.The training and evaluation are carried out in the constructed surveillance video feature conflict dataset,and eight scenes are constructed in the way of orthogonal experiments.The experimental results show that the performance of object detection can be significantly improved by using the proposed method in the object detection task of the intelligent transportation system surveillance video feature conflict scene.
基金supported by the National Natural Science Foundation of China(82174513)Fundamental Research Funds for the Central Public Welfare Research Institutes(ZZ-JQ2023003).
文摘To systematically elucidate the central role of the cardiac autonomic nervous system(ANS)in maintaining cardiovascular homeostasis,analyze the pathological mechanisms underlying its dysregulation,integrate multidisciplinary research findings from the U.S.stimulating peripheral nerves to alleviate disease symptoms(SPARC)program,and evaluate the unique advantages and current limitations of acupuncture in modulating cardiac ANS function,ultimately proposing novel strategies for the prevention and treatment of cardiovascular diseases.A systematic literature review was conducted to synthesize current knowledge on the fundamental regulatory mechanisms of the cardiac ANS,the SPARC program's innovative contributions to neuroanatomy and neural pathway mapping,as well as clinical and experimental evidence supporting acupuncture's modulation of the ANS.The SPARC program has made significant progress in elucidating the anatomical organization and neural circuitry of the cardiac ANS through interdisciplinary collaboration,offering novel insights and methodological frameworks for studying cardiac autonomic regulation.Acupuncture,as a cornerstone of traditional medicine,has demonstrated both specificity and multi-target regulatory effects on the cardiac ANS through clinical and experimental studies.However,challenges remain,including an incomplete mechanistic understanding,technical limitations in research methodologies,and difficulties in translating findings into clinical practice.Future research on acupuncture should build upon its inherent strengths while advancing its integration with modern scientific and technological approaches.Strengthening interdisciplinary collaboration and leveraging artificial intelligence can open new frontiers in mechanistic exploration and technological innovation.These efforts will facilitate the internationalization of acupuncture research and contribute innovative perspectives and therapeutic strategies for cardiovascular disease prevention and treatment.
基金funded by the Open Access Initiative of the University of Bremen and the DFG via SuUB BremenThe authors extend their appreciation to the Deanship of Research and Graduate Studies at King Khalid University for funding this work through Large Group Project under grant number(RGP2/367/46)+1 种基金This research is supported and funded by Princess Nourah bint Abdulrahman University Researchers Supporting Project number(PNURSP2025R410)Princess Nourah bint Abdulrahman University,Riyadh,Saudi Arabia.
文摘As urban landscapes evolve and vehicular volumes soar,traditional traffic monitoring systems struggle to scale,often failing under the complexities of dense,dynamic,and occluded environments.This paper introduces a novel,unified deep learning framework for vehicle detection,tracking,counting,and classification in aerial imagery designed explicitly for modern smart city infrastructure demands.Our approach begins with adaptive histogram equalization to optimize aerial image clarity,followed by a cutting-edge scene parsing technique using Mask2Former,enabling robust segmentation even in visually congested settings.Vehicle detection leverages the latest YOLOv11 architecture,delivering superior accuracy in aerial contexts by addressing occlusion,scale variance,and fine-grained object differentiation.We incorporate the highly efficient ByteTrack algorithm for tracking,enabling seamless identity preservation across frames.Vehicle counting is achieved through an unsupervised DBSCAN-based method,ensuring adaptability to varying traffic densities.We further introduce a hybrid feature extraction module combining Convolutional Neural Networks(CNNs)with Zernike Moments,capturing both deep semantic and geometric signatures of vehicles.The final classification is powered by NASNet,a neural architecture search-optimized model,ensuring high accuracy across diverse vehicle types and orientations.Extensive evaluations of the VAID benchmark dataset demonstrate the system’s outstanding performance,achieving 96%detection,94%tracking,and 96.4%classification accuracy.On the UAVDT dataset,the system attains 95%detection,93%tracking,and 95%classification accuracy,confirming its robustness across diverse aerial traffic scenarios.These results establish new benchmarks in aerial traffic analysis and validate the framework’s scalability,making it a powerful and adaptable solution for next-generation intelligent transportation systems and urban surveillance.
文摘This survey presents a comprehensive examination of sensor fusion research spanning four decades,tracing the methodological evolution,application domains,and alignment with classical hierarchical models.Building on this long-term trajectory,the foundational approaches such as probabilistic inference,early neural networks,rulebasedmethods,and feature-level fusion established the principles of uncertainty handling andmulti-sensor integration in the 1990s.The fusion methods of 2000s marked the consolidation of these ideas through advanced Kalman and particle filtering,Bayesian–Dempster–Shafer hybrids,distributed consensus algorithms,and machine learning ensembles for more robust and domain-specific implementations.From 2011 to 2020,the widespread adoption of deep learning transformed the field driving some major breakthroughs in the autonomous vehicles domain.A key contribution of this work is the assessment of contemporary methods against the JDL model,revealing gaps at higher levels-especially in situation and impact assessment.Contemporary methods offer only limited implementation of higher-level fusion.The survey also reviews the benchmark multi-sensor datasets,noting their role in advancing the field while identifying major shortcomings like the lack of domain diversity and hierarchical coverage.By synthesizing developments across decades and paradigms,this survey provides both a historical narrative and a forward-looking perspective.It highlights unresolved challenges in transparency,scalability,robustness,and trustworthiness,while identifying emerging paradigms such as neuromorphic fusion and explainable AI as promising directions.This paves the way forward for advancing sensor fusion towards transparent and adaptive next-generation autonomous systems.
基金supported by Korea Institute of Police Technology(KIPoT)grant funded by the Korea government(KNPA)(Project Name:Development of Lv.4 Driving Ability Evaluation Technology for Autonomous Vehicles Based on Real Roads/Project Number:RS-2023-00238253).
文摘Analyzing the driving behavior of autonomous vehicles(AV)in mixed traffic conditions at urban intersections has become increasingly important for improving intersection design,providing infrastructure-based guidance information,and developing capability-enhanced AV perception systems.This study investigated the contributing factors affecting AV driving behavior using theWaymo Open Dataset.Binarized autonomous driving stability metrics,derived via a kernel density estimation,served as the target variables for a random forest classification model.The model’s input variables included 15 factors divided into four types:intersection-related,surrounding object-related,road infrastructure-related,and time-of-day-related types.The random forest classification model was employed to identify the key factors affecting autonomous driving behavior.In addition,the identified factors were further ranked based on feature importance.SHAP analysis was utilized to enhance model interpretability by quantifying the contribution of each factor and identifying their directional impacts.The type of intersection factor was found to have an importance of 0.243 and was the most influential factor on autonomous driving behavior.On average,intersection-related factors had an importance of 0.196,which is approximately a 31.1%margin over the average importance of surrounding object-related factors.Additionally,the surrounding object-related factors that were collected through sensors on the autonomous vehicle had a high degree of feature importance,especially with the number of pedestrians having the highest importance(0.107)of the types of objects.The correlation between these findings can contribute to the development of various treatments to improvemore harmonized AVs’maneuvering with other road users and facilities in urban mixed traffic environments.
文摘This study aimed to enhance the performance of semantic segmentation for autonomous driving by improving the 2DPASS model.Two novel improvements were proposed and implemented in this paper:dynamically adjusting the loss function ratio and integrating an attention mechanism(CBAM).First,the loss function weights were adjusted dynamically.The grid search method is used for deciding the best ratio of 7:3.It gives greater emphasis to the cross-entropy loss,which resulted in better segmentation performance.Second,CBAM was applied at different layers of the 2Dencoder.Heatmap analysis revealed that introducing it after the second block of 2D image encoding produced the most effective enhancement of important feature representation.The training epoch was chosen for optimizing the best value by experiments,which improved model convergence and overall accuracy.To evaluate the proposed approach,experiments were conducted based on the SemanticKITTI database.The results showed that the improved model achieved higher segmentation accuracy by 64.31%,improved 11.47% in mIoU compared with the conventional 2DPASS model(baseline:52.84%).It was more effective at detecting small and distant objects and clearly identifying boundaries between different classes.Issues such as noise and variations in data distribution affected its accuracy,indicating the need for further refinement.Overall,the proposed improvements to the 2DPASS model demonstrated the potential to advance semantic segmentation technology and contributed to a more reliable perception of complex,dynamic environments in autonomous vehicles.Accurate segmentation enhances the vehicle’s ability to distinguish different objects,and this improvement directly supports safer navigation,robust decision-making,and efficient path planning,making it highly applicable to real-world deployment of autonomous systems in urban and highway settings.
