Event cameras have emerged as a promising sensor for object detection in high-speed and high-dynamic-range scenarios. However, existing methods often struggle with the inherent spatiotemporal sparsity and noise of event data. In this paper, we propose a novel framework AlCo that integrates a Semantic-Geometry Al ignment Convolution (SGAC) and a Hierarchical Semantic Co mpensation (HSC) block. SGAC explicitly models geometric priors and semantic alignment to suppress noise and enhance structural consistency in early-stage features. The HSC block further employs a dual-branch strategy: a sparse branch with prototype-guided token selection and diversity-aware recycling to maintain efficiency, and a global branch to compensate for long-range contextual information. Extensive experiments on the Gen1 and 1Mpx datasets demonstrate that our method achieves state-of-the-art performance while providing favorable FLOPs, latency, and memory trade-offs.

Infrared and visible image fusion (IVIF) aims to extract fine details from visible images and complementary information from infrared images. Most existing methods directly extract relevant and complementary features from each modality using neural networks, often overlooking the guidance process and the distinct frequency-domain characteristics of these features. To address this, we propose HRFusion-a novel frequency-domain framework that extracts complementary features from hybrid features using prior-constrained relevant features, effectively enhancing complementary information and reducing redundancy. In HRFusion, hybrid and relevant features are robustly extracted to guide the subsequent fusion stage. By leveraging frequency differences between complementary and relevant features, we introduce the Enhanced Complementary Frequency Network (ECFNet), which uses optimized Variational Mode Decomposition (VMD) to effectively separate and process these signals for fusion. The overall architecture is built with the proposed DTBlock, which captures both global and local features. Extensive experiments show that our method achieves state-of-the-art performance on the TNO, MSRS, M3FD, and Harvard Brain datasets, outperforming recent approaches. Code is available at https://github.com/liuuuuu777/HRFusion.

Monocular depth estimation stands as a fundamental pursuit in computer vision. Recently, some methods have attempted to introduce the text-to-image diffusion model into the domain of monocular depth estimation and achieved impressive results. However, these methods typically employ pre-defined templates as text prompts to guide the learning of denoising networks, resulting in limited flexibility and scalability. In this paper, we propose OGDepth, a diffusion-based monocular depth estimation network with object prompts generated by taking advantage of the object detection information from the scene. Specifically, we design an Object Prompt Module (OPM) to encode the object detection information into prompts that are more closely aligned with the image content, offering richer contextual information while circumventing the monotony and redundancy inherent in template-generated prompts. Moreover, we employ bounding box information for each object to filter and localize objects, enabling the model to grasp relative positional information within the scene. This facilitates the creation of a more precise depth map. Additionally, we design a Global-Local Interaction Decoder (GLID) to facilitate the mutual exchange of features at different scales, enabling efficient feature fusion. Our approach underwent rigorous experiments across multiple datasets, with results showcasing its state-of-the-art performance. Notably, on the KITTI dataset, our model achieves an RMSE of 1.967 and a REL of 0.047, and both metrics are the best among all compared methods. On the NYU Depth V2 dataset, our method achieves an RMSE score of 0.221, representing a notable 12.9% enhancement compared to the baseline method (VPD).

Behavior Trees (BTs) is a robust framework for decision-making that is highly applicable in automated guided vehicles (AGVs), offering a way to manage complex tasks and respond to dynamic changes in their environment.To enhance the adaptability and efficiency of BTs in AGVs, we introduce a hybrid approach that integrates Genetic Programming (GP) and Reinforcement Learning (RL).The GP evolves the BT structure by selecting potential actions from an action pool, guided by tailored constraints that ensure the trees remain interpretable and relevant to AGV tasks.Meanwhile, to mitigate node dependency issues in BTs, we employ RL, which incorporates a parameterdependent dynamic updating (PDDU) algorithm to monitor and manage the relationships between parameters.Furthermore, we implement a weighted ϵ-greedy algorithm to refine the parameter update process.Our methodology is validated through simulated AGV scenarios, demonstrating that the evolved BT significantly improve AGV autonomy.This innovative fusion of GP and RL techniques sets a foundation for future developments in AGV technology, with potential applications extending beyond the factory floor to any environment where AGVs are deployed.

With the rapid advances in high-speed mobile networks such as 5G, Wi-Fi 6, and the upcoming 6G and Wi-Fi 7, streaming live video has become ubiquitous for mobile users. However, live video is susceptible to short-term network condition fluctuations which could lead to video stalls. Our investigations revealed that a substantial portion of such fluctuations were in fact caused by RTT spikes that were not congestion-related. These often confuse the transport protocol into dropping the transmission rate significantly, resulting in video stalls. This motivated us to develop a novel scheme called CWnd-loan to reduce the sender's CWnd-limited idle time during RTT spikes. We applied CWnd-loan to the QUIC protocol with BBR/CUBIC congestion control and strategically deployed it in a tier-1 live video service. The results show that CWnd-loan can effectively reduce sender CWnd-limited idle time by up to 18%, consequently reducing the duration and number of live video stalls by as much as 8.9% and 10.6%. Furthermore, CWnd-loan can also reduce the first-frame time and the playback failure rate by up to 3.2% and 2.7%, respectively. CWnd-loan is designed to complement existing congestion control algorithms and thus could potentially be applied to current as well as future TCP/QUIC designs to tackle RTT spikes commonly found across mobile and wireless networks.

Learning-based Internet congestion control algorithms have attracted much attention due to their potential performance improvement over traditional algorithms. However, such performance improvement is usually at the expense of black-box design and high computational overhead, which prevent them from large-scale deployment over production networks. To address this problem, we propose a novel Internet congestion control algorithm called Gemini. It contains a parameterized congestion control module, which is white-box designed with low computational overhead, and an online parameter optimization module, which serves to adapt the parameterized congestion control module to different networks for higher transmission performance. Extensive trace-driven emulations reveal Gemini achieves better balances between delay and throughput than state-of-the-art algorithms. Moreover, we successfully deploy Gemini over production networks. The evaluation results show that the average throughput of Gemini is 5% higher than that of Cubic (4% higher than that of BBR) over a mobile application downloading service and 61% higher than that of Cubic (33% higher than that of BBR) over a commercial network speed-test benchmarking service.

The robust adaptive beamforming (RAB) problem is considered via the worst-case signal-to-interference-plus-noise ratio (SINR) maximization over distributional uncertainty sets for the random interference-plus-noise covariance (INC) matrix and desired signal steering vector. The distributional uncertainty set of the INC matrix accounts for the support and the positive semidefinite (PSD) mean of the distribution, and a similarity constraint on the mean. The distributional uncertainty set for the steering vector consists of the constraints on the known first- and second-order moments. The RAB problem is formulated as a minimization of the worst-case expected value of the SINR denominator achieved by any distribution, subject to the expected value of the numerator being greater than or equal to one for each distribution. Resorting to the strong duality of linear conic programming, such a RAB problem is rewritten as a quadratic matrix inequality problem. It is then tackled by iteratively solving a sequence of linear matrix inequality relaxation problems with the penalty term on the rank-one PSD matrix constraint. To validate the results, simulation examples are presented, and they demonstrate the improved performance of the proposed robust beamformer in terms of the array output SINR.

With the progressive deployment of 5G networks around the world, mobile networks are entering a new era where bandwidth will be breaking through the Gbps barrier. In this work, we investigate the performance of current TCP designs in such high-bandwidth networks, demonstrating the potential bottleneck due to TCP’s Slow-Start mechanism which is an integral component in most TCP designs. For example, transferring a file of 1 MB size in a first-generation 5G network using Linux’s default TCP-Cubic and Google’s TCP-BBR resulted in average throughputs of 18.2 Mbps and 32.8 Mbps, respectively. Compared to the mean available bandwidth of 180 Mbps, the gap is significant. To tackle this problem, we developed an enhanced Stateful-TCP technique to transform BBR into a new S-BBR to accelerate its startup performance to narrow the gap. Results from trace-driven emulated 5G network experiments show that S-BBR could improve BBR’s throughput performance by 50% to 100% while maintaining similar delay performance. This is further validated by an independent competitive benchmark using over 500 clients where S-BBR raised BBR’s throughput by 69%. S-BBR is sender-based and thus can be readily deployed in Internet servers without any requirements from the client side, it retains BBR’s desirable features and so offers a promising solution to enhance mobile applications’ performance in the emerging high-bandwidth mobile and wireless networks.

Campus networks are widely deployed for organizations like universities and large companies. Applications and network-based services require campus networks to guarantee short queue and provide low latency and large bandwidth. However, the widely adopted packet-loss-based congestion control mechanism in client hosts builds up long queues in the switch buffer, which is prone to packet loss in burst scenarios, resulting in great network delay. Therefore, a scheme for efficiently controlling queue length of shallow buffer switches in campus networks is urgently needed. Explicit Congestion Notification(ECN) as an explicit feedback mechanism is widely adopted in data center networks to build lossless networks. In this paper, we propose Clean, an efficient queue length control scheme based on transparent ECN-proxy for campus networks. Clean is able to exert fine-grained control over arbitrary client TCP stacks by enforcing per-flow congestion control in the access point(AP). It allows the campus network switches to maintain a low queue length, resulting in high throughput, low latency and zero packet loss. Evaluation results demonstrate that Clean reduces the maximum queue length of the switch by 86% and reduces the 99th percentile latency by 85%. Clean also achieves zero packet loss in burst scenarios.