While cooperative perception can overcome the limitations of single-vehicle systems, the practical implementation of vehicle-to-vehicle and vehicle-to-infrastructure systems is often impeded by significant economic barriers. Aerial-ground cooperation (AGC), which pairs ground vehicles with drones, presents a more economically viable and rapidly deployable alternative. However, this emerging field has been held back by a critical lack of high-quality public datasets and benchmarks. To bridge this gap, we present Griffin, a comprehensive AGC 3D perception dataset, featuring over 250 dynamic scenes (37k+ frames). It incorporates varied drone altitudes (20-60m), diverse weather conditions, realistic drone dynamics via CARLA-AirSim co-simulation, and critical occlusion-aware 3D annotations. Accompanying the dataset is a unified benchmarking framework for cooperative detection and tracking, with protocols to evaluate communication efficiency, altitude adaptability, and robustness to communication latency, data loss and localization noise. By experiments through different cooperative paradigms, we demonstrate the effectiveness and limitations of current methods and provide crucial insights for future research.

Generative AI for automated glaucoma diagnostic report generation faces two predominant challenges: content redundancy in narrative outputs and inadequate highlighting of pathologically significant features including optic disc cupping, retinal nerve fiber layer defects, and visual field abnormalities. These limitations primarily stem from current multimodal architectures' insufficient capacity to extract discriminative structuraltextural patterns from fundus imaging data while maintaining precise semantic alignment with domain-specific terminology in comprehensive clinical reports. To overcome these constraints, we present the Dual-Attention Semantic Parallel-LSTM Network (DA-SPL), an advanced multimodal generation framework that synergistically processes both fundus imaging and supplementary visual inputs. DA-SPL employs an Encoder-Decoder structure augmented with the novel joint dual-attention mechanism in the encoder for cross-modal feature refinement, the parallelized LSTM decoder architecture for enhanced temporal-semantic consistency, and the specialized label enhancement module for accurate disease-relevant term generation. Rigorous evaluation on standard glaucoma datasets demonstrates DA-SPL's consistent superiority over state-of-the-art models across quantitative metrics. DA-SPL exhibits exceptional capability in extracting subtle pathological indicators from multimodal inputs while generating diagnostically precise reports that exhibit strong concordance with clinical expert annotations. Code is here: https://github.com/CH-YellowOrange/DA-SPL.

Since 2023, Vector Quantization (VQ)-based discrete generation methods have rapidly dominated human motion generation, primarily surpassing diffusion-based continuous generation methods in standard performance metrics. However, VQ-based methods have inherent limitations. Representing continuous motion data as limited discrete tokens leads to inevitable information loss, reduces the diversity of generated motions, and restricts their ability to function effectively as motion priors or generation guidance. In contrast, the continuous space generation nature of diffusion-based methods makes them well-suited to address these limitations and with even potential for model scalability. In this work, we systematically investigate why current VQ-based methods perform well and explore the limitations of existing diffusion-based methods from the perspective of motion data representation and distribution. Drawing on these insights, we preserve the inherent strengths of a diffusion-based human motion generation model and gradually optimize it with inspiration from VQ-based approaches. Our approach introduces a human motion diffusion model enabled to perform masked autoregression, optimized with a reformed data representation and distribution. Additionally, we propose a more robust evaluation method to assess different approaches. Extensive experiments on various datasets demonstrate our method outperforms previous methods and achieves state-of-the-art performances.

Safety is one of the most crucial challenges of autonomous driving vehicles, and one solution to guarantee safety is to employ an additional control revision module after the planning backbone. Control Barrier Function (CBF) has been widely used because of its strong mathematical foundation on safety. However, the incompatibility with heterogeneous perception data and incomplete consideration of traffic scene elements make existing systems hard to be applied in dynamic and complex real-world scenarios. In this study, we introduce a generalized control revision method for autonomous driving safety, which adopts both vectorized perception and occupancy grid map as inputs and comprehensively models multiple types of traffic scene constraints based on a new proposed barrier function. Traffic elements are integrated into one unified framework, decoupled from specific scenario settings or rules. Experiments on CARLA, SUMO, and OnSite simulator prove that the proposed algorithm could realize safe control revision under complicated scenes, adapting to various planning backbones, road topologies, and risk types. Physical platform validation also verifies the real-world application feasibility.

Zero-shot referring expression comprehension aims at localizing bounding boxes in an image corresponding to provided textual prompts, which requires: (i) afine-grained disentanglement of complex visual scene and textual context, and (ii) a capacity to understand relationships among disentangled entities. Unfortunately, existing large vision-language alignment (VLA) models, e.g., CLIP, struggle with both aspects so cannot be directly used for this task. To mitigate this gap, we leverage large foundation models to disentangle both images and texts into triplets in the for-mat of (subject, predicate, object). After that, grounding is accomplished by calculating the structural similarity matrix between visual and textual triplets with a VLA model, and subsequently propagate it to an instance-level similarity matrix. Furthermore, to equip VLA models with the ability of relationship understanding, we design a triplet-matching objective to fine-tune the VLA models on a collection of curated dataset containing abundant entity relationships. Experiments demonstrate that our visual grounding performance increase of up to 19.5% over the SOTA zero-shot model on RefCOCO/+/g. On the more challenging Who's Waldo dataset, our zero-shot approach achieves comparable accuracy to the fully super-vised model. Code is available at https://github.com/Show-han/Zeroshot_REC.

Logo classification systems have become increasingly important in various industries for tasks, such as infringement detection and industrial production. However, challenges still exist in logo classification due to real-world image background interference, the high similarity between classes, labeling difficulties, and the insufficient representation of occlusion in single-view logos. Many existing algorithms fail to consider the data characteristics and the intrinsic information of multiple views, which limits their performance. To overcome these limitations, we developed a novel Cross-View Information Awareness Network (CVIA-Net) for logo classification. To differentiate between similar logo categories, the CVIA-Net novel learns context-shared features of the same category via a self-supervised way without labeled, which solves the problem of insufficient features due to occlusion. For single-view images, CVIA-Net establishes a “bottleneck” representation to address background interference. Extensive experiments on three datasets demonstrate that it outperforms state-of-the-art methods. The method is expected to advance the development of cross-view representation learning.

Global path planning is an essential component of intelligent vehicle study. This paper designs a two-layer global path planning method based on an improved A* algorithm and quadratic programming for UGVs in a large-scale off-road environment. In the first layer, we generate a global path from the current node to the target node via an improved A* algorithm, with a grid map containing the information of non-accessible areas and uncertainty of off-road terrains as input. The second layer smooths the entire path based on quadratic programming. We adopt efficiency improvement methods in both layers, which ensure the real-time performance of the algorithm. The planner has been verified by simulation and experiments, and the results validate the practicability and real-time performance of the designed method.

In this paper, we consider asynchronous massive connectivity, where massive low-power and low-rate devices with sporadic activity patterns connect to a multi-antenna access point (AP) in an asynchronous manner. Asynchronous transmission minimizes the amount of coordination between the devices and the AP, and thereby simplifies the transmitter design, yet at the cost of a more challenging receiver design. Specifically, asynchronous transmission results in inter-symbol interference (ISI) since the sampling at AP generally cannot match with the symbol intervals of uncoordinated devices. To enable reliable reception, we develop a turbo approximate message passing (TAMP) algorithm that consists of a channel-signal decomposition (CSD) module and a delay learning (DL) module. The CSD carries out sparse matrix factorization to estimate the channels and the ISI corrupted signals of active devices, and the DL is designed to estimate the delay of each active user and resolve the corresponding ISI based on the Bayesian principle. To refine the delay estimation, we further divide the DL module into symbol-level delay learning (SDL) and sub-symbol-level delay learning (sub-SDL) submodules. In particular, the sub-SDL estimates the residue delays (obtained by taking modulo of the symbol interval) and then finely compensates the ISI. Due to the continuity and randomness of time delay, the receive signal constellation consists of lines and curves instead of discrete points, even if the transmit signal constellation is discrete. To reduce the complexity of soft demodulation, we introduce a truncation and projection based approximation method to simplify the related message calculation. Numerical results demonstrate the superior performance of the proposed TAMP algorithm. Particularly, the TAMP algorithm is able to approach the single-user bound with known user delay.

In this paper, we develop an asynchronous massive access framework. Besides the multiple-access interference issue, asynchronous transmission induces inter-symbol interference effect of the same user that is determined by the unknown user delay. To solve this problem, we propose an Bayesian receiver where the whole receiver is divided into two modules: the channel-signal decomposition (CSD) module and the delay learning (DL) module. The CSD module demixes the transmit signals of different users by leveraging the bilinear generalized approximate message passing (BiGAMP) algorithm, and the DL module is designed to estimate the time delay of each user based on the Bayesian principle. Additionally, due to the continuity of time delay, the constellation of all possible received user signals consists of lines and curves instead of discrete points, even if the original transmit signals are discrete. To reduce complexity, we introduce a truncation and projection based approximation method to simplify the related message calculation. Numerical results demonstrate the superior performance of the proposed scheme. Particularly, the proposed scheme is able to approach the single-user interference-free bound with known user delay.

This work investigates the synchronization problem of a dynamical network composed of nonlinear nodes with directed couplings and suffering from heterogeneous delays. Specifically, a time-varying pinning control is adopted to achieve synchronization. The system’s dynamics are interpreted under a multiplex network framework, from which an indicator is designed to measure the synchronizability of the pinned network. Criteria are then derived to ensure exponential synchronization while the largest admissible coupling delay is also estimated. These criteria are based on the time average of the proposed indicator over some user-defined time intervals. Therefore, it possesses high versatility to support a wide range of scenarios with different intermittent control strategies. An interesting observation is that the proposed indicator is always upper bounded, manifesting that if the couplings between the nodes are too weak, pinning synchronization may fail no matter how large the pinning gain is. To further demonstrate the applicability of the theorem and the corollaries, numerical examples under different pinning configurations are presented.

In this work, the consensus problem of networked agents under denial-of-service (DoS) attacks is studied. Due to DoS attacks, communication channels would be intermittently congested, which inevitably affects the state update in the consensus process and leads to the failure of a consensus protocol. Such a situation has been recast into a consensus problem with on-off control, as suggested by other studies. However, it is commonly assumed that the underlying graph is undirected and the DoS attack follows certain periodic patterns. Obviously, these assumptions cannot truly reflect some real cases. Therefore, aperiodic attacks and directed topology are considered here. Our theoretical results reveal that consensus is achievable, provided that the average of the DoS attack duration over a certain length of time interval is upper bounded. Compared with other existing results, this new consensus criterion demonstrates better generality and tractability. Finally, simulations are presented to further demonstrate the effectiveness of the theoretical results.

As one of the promising initiatives for environmental preservation, electric vehicles (EVs) have been deploying in many countries. However, the limited drivable range of EV usually necessitates mid-way re-charging for long-distance journey, which is one of the most concerning issues in EV driving. Given limited public charging resources, EV drivers easily suffer from serious “range anxiety”, in particular when the queuing time at the charging station is long. Therefore, it is important to distribute the EV flows to the right charging stations for recharging, while at the same time, to minimize the overall journey time. In this paper, such an issue is formulated as an optimization problem and solved by a dedicated flow distribution algorithm (FDA). The practical usage of the design is demonstrated by extensive simulations upon the eastern Massachusetts highway network.