KTKang Tan

Kang Tan

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Kang Tan has a total of 42 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    PA-YOLO: Partial-Channel Grouped Attention and Multi-Scale Prediction for Object Detection in UAV Imagery

    by Kang Tan, Hu Yao, Zhiyuan Liu, Xingjie Zhang, Qiang Zhang, Xiao Wu, Xinggan Peng, Xingzhou Chen

    Unmanned aerial vehicle (UAV) imagery presents small-object footprints, dense layouts, occlusion, and weak appearance cues, requiring both local spatial detail and broader context under constrained computation. Existing high-resolution prediction paths and attention mechanisms can strengthen these representations, but their extensive use increases feature-map, attention, and fusion costs. We propose Partial Attention–YOLO (PA-YOLO); its primary methodological contribution is a partial attention module with multi-group query attention, which applies grouped contextual modeling to selected channels while retaining a direct bypass for locally derived convolutional detail. The architecture further integrates established RepVGG-style re-parameterized downsampling with high-resolution four-scale feature fusion and detection. Fixed-recipe component ablation on VisDrone indicates complementary contributions from partial-channel attention and the integrated multi-scale design. On VisDrone, PA-YOLO-n and PA-YOLO-s report AP50/AP50:95 point estimates of 39.4/23.8% and 45.0/27.2%, respectively; under a secondary, non-official 2000-frame UAVDT-derived protocol, the corresponding values are 96.6/69.3% and 97.3/72.0%. These accuracies are single-run point estimates. Controlled forward-only profiling on one RTX 3090 platform gives mean eager-FP32 latencies of 13.795 and 14.197 ms/image, showing that compact parameterization does not necessarily yield lower inference latency than convolution-centric baselines. These findings motivate deployment-oriented optimization and onboard evaluation.

  • Open Access

    A Partial-Attention and Multi-Scale Fusion Framework for Efficient Object Detection in Unmanned Aerial Vehicle Imagery

    by Kang Tan, Hu Yao, Zhiyuan Liu, Xingjie Zhang, Qiang Zhang, Xiao Wu, Xingan Peng, Xingzhou Chen, Ce Zhu

    Unmanned aerial vehicles (UAVs) are crucial platforms for visual perception tasks in smart cities, including traffic monitoring and vehicle tracking. Deep learning-based object detection in UAV imagery remains challenging due to complex backgrounds, small object sizes, dense target distributions, and limited onboard computational resources. This paper introduces Partial Attention–You Only Look Once (PA-YOLO), a lightweight yet high-performing deep-learning detection framework designed for UAV imagery. The proposed architecture introduces three key innovations: (1) a partial attention module that selectively applies attention to a subset of feature channels, enhancing global contextual representation while maintaining computational efficiency; (2) a re-parameterized convolutional backbone based on the Visual Geometry Group (VGG) design (RepVGG), which boosts feature diversity during training and allows efficient single-branch inference after deployment; and (3) a grouped multi-scale detection head that incorporates an additional high-resolution branch and grouped convolutions to improve small-object detection and robustness under occlusion. Extensive experiments on the VisDrone and UAV Detection and Tracking (UAVDT) benchmarks demonstrate that PA-YOLO achieves a superior accuracy–parameter trade-off compared to state-of-the-art lightweight detectors, achieving a mean average precision (mAP) of 0.450 at an intersection over union (IoU) threshold of 0.5 while maintaining a compact model size. This corroborates that PA-YOLO effectively balances detection accuracy with model compactness, rendering it suitable for real-world UAV applications in intelligent transportation and aerial surveillance.

  • Multimodal Sensing for Intelligent V2X: A Review of Recent Advances Toward Deployment

    by Kang Tan, Ce Zhu

    The integration of multimodal sensing and communication systems advances the integrated sensing and communication (ISAC) concept by combining camera, LiDAR, radar, and radio-frequency (RF) data. This approach provides a promising avenue to enhance wireless link reliability and spectral efficiency while reducing end-to-end latency in 6G vehicle-to-everything (V2X) networks. Surveys on sensing and communication integration primarily concentrate on RF technologies. While initial studies have explored the combination of various sensing methods, there is a paucity of comprehensive reviews regarding how such data is utilized to enhance V2X communication performance. Therefore, this review comprehensively analyzes recent research on multimodal sensing-assisted vehicular networks to enhance communication performance, centered around channel prediction, beam management, mobility management, and multitask solutions. Since data-centric techniques are the main approach in this field, relevant datasets and simulators supporting data-driven designs are also thoroughly examined. Finally, by critically analyzing current research, the survey highlights unresolved issues and future trends in research cornerstones, method implementation, and system design. It also addresses key deployment challenges—spectrum availability, roadside hardware costs, and privacy-aware learning—thereby charting the course for developing real-world intelligent V2X networks.


  • Multi-Agent Deep Reinforcement Learning for Vehicular Resource Allocation: A Comparison Study of Different Agent Cooperation Levels

    by Kang Tan, Ce Zhu

    The spectrum sharing design (such as the underlay sidelink configuration in cellular networks) for vehicle-to-vehicle (V2V) and vehicle-to-network (V2N) links can improve the efficiency of spectrum utilization for vehicular networks. However, differentiated service requirements for V2V and V2N links and the greater complexity of interference management make it challenging for conventional centralized learning-based methods in the face of real-world deployment. This motivates the exploration of multi-agent deep reinforcement learning (DRL) solutions, while existing research has predominantly focused on cooperative setting that causes large communication overheads for real-world deployment. This paper proposed a decentralized resource allocation scheme for joint spectrum sub-band and V2V transmission power selection that utilizes federated learning (FL) to formulate a mixed setup to avoid frequent communication and enable easy model sharing, termed FL-MX for the design of FL and mixed DRL agent cooperation level. Different levels of cooperation among agents are also explored by implementing different state-of-the-art (SOTA) DRL-based resource allocation algorithms on the developed standard-compliant simulation platform. Through extensive simulations, it is observed that the proposed FL-MX manages to converge effectively and can match the best SOTA performance, gaining a 22.55% and 9.5% improvement in the V2V payload delivery ratio and V2N throughput compared to the random baseline. Such results make it promising for future research to investigate multi-agent DRL designs at a mixed agent cooperation level for high-performing network decision solutions with less frequent communication demands.


  • Open Access

    Location estimation for supporting adaptive beamforming

    by Jaspreet Kaur, Kang Tan, Arslan Shafique, Olaoluwa R. Popoola, Muhammad A. Imran, Qammer H. Abbasi, Hasan T. Abbas

  • Open Access

    Fingerprinting‐Based Indoor Localization in a 3 × 3 Meter Grid Using OFDM Signals at Sub‐6 GHz

    by Jaspreet Kaur, Kang Tan, Muhammad Z. Khan, Olaoluwa R. Popoola, Muhammad A. Imran, Qammer H. Abbasi, Hasan T. Abbas

    Accurately determining the indoor location of mobile devices has garnered significant interest due to the complex challenges posed by non‐line‐of‐sight (NLOS) propagation and multipath effects. To address this challenge, this paper proposes a new approach to indoor positioning that utilises channel state information (CSI) and machine learning (ML) techniques to improve accuracy. The proposed method extracts the amplitude and phase differences of the subcarriers from the CSI data to create fingerprints. ML algorithms and network architecture are utilised to train the CSI data from two antennas, in the form of phase and amplitude. Experiments conducted in a standard indoor environment demonstrate the effectiveness of the proposed method.


  • Meta-Transfer Learning-Based Handover Optimization for V2N Communication

    by Rana Muhammad Sohaib, Oluwakayode Onireti, Kang Tan, Yusuf Sambo, Rafiq Swash, Muhammad Ali Imran

    The rapid growth of vehicle-to-network (V2N) communication demands efficient handover decision-making strategies to ensure seamless connectivity and maximum throughput. However, the dynamic nature of V2N scenarios poses challenges for traditional handover algorithms. To address this, we propose a deep reinforcement learning (DRL)-based approach for optimizing handover decisions in dynamic V2N communication. We leverages the advantages of transfer learning and meta-learning to generalize across time-evolving source and target tasks. In this paper, we derive generalization bounds for our DRLbased approach, specifically focusing on optimizing the handover process in V2N communication. The derived bounds provide theoretical guarantees on the expected generalization error of the learned handover time function for the target task. To implement our framework, we propose a meta-learning framework, Adaptto-evolve (A2E), based on the double deep Q-networks (DDQN) with Thompson sampling approach. The A2E framework enables quick adaptation to new tasks by minimizing the error upper bounds with divergence measures. Through transfer learning, the meta-learner dynamically evolves its handover decision-making strategy to maximize average throughput while reducing the number of handovers. We use Thompson sampling with the DDQN to balance exploration and exploitation. The DDQN with The Thompson sampling approach, ensuring efficient and effective learning, forms the foundation for optimizing the metatraining process, resulting in improvement in cumulated packet lossby48.02%in highway settings and 46.32%in rural settings


  • Open Access

    Contextual beamforming: Exploiting location and AI for enhanced wireless telecommunication performance

    by Jaspreet Kaur, Satyam Bhatti, Kang Tan, Olaoluwa R. Popoola, Muhammad Ali Imran, Rami Ghannam, Qammer H. Abbasi, Hasan T. Abbas

    Beamforming, an integral component of modern mobile networks, enables spatial selectivity and improves network quality. However, many beamforming techniques are iterative, introducing unwanted latency to the system. In recent times, there has been a growing interest in leveraging mobile users’ location information to expedite beamforming processes. This paper explores the concept of contextual beamforming, discussing its advantages, disadvantages, and implications. Notably, we demonstrate an impressive 53% improvement in the signal-to-interference-plus-noise ratio by implementing the adaptive beamforming maximum ratio transmission (MRT) algorithm compared to scenarios without beamforming. It further elucidates how MRT contributes to contextual beamforming. The importance of localization in implementing contextual beamforming is also examined. Additionally, the paper delves into the use of artificial intelligence (AI) schemes, including machine learning and deep learning, in implementing contextual beamforming techniques that leverage user location information. Based on the comprehensive review, the results suggest that the combination of MRT and zero-forcing techniques, alongside deep neural networks employing Bayesian optimization, represents the most promising approach for contextual beamforming. Furthermore, the study discusses the future potential of programmable switches, such as Tofino—an innovative switch developed by Barefoot Networks (now a part of Intel)—in enabling location-aware beamforming. This paper highlights the significance of contextual beamforming for improving wireless telecommunications performance. By capitalizing on location information and employing advanced AI techniques, the field can overcome challenges and unlock new possibilities for delivering reliable and efficient mobile networks.


  • Open Access

    Graph neural network-based cell switching for energy optimization in ultra-dense heterogeneous networks

    by Kang Tan, Duncan Bremner, Julien Le Kernec, Yusuf Sambo, Lei Zhang, Muhammad Ali Imran

    The development of ultra-dense heterogeneous networks (HetNets) will cause a significant rise in energy consumption with large-scale base station (BS) deployments, requiring cellular networks to be more energy efficient to reduce operational expense and promote sustainability. Cell switching is an effective method to achieve the energy efficiency goals, but traditional heuristic cell switching algorithms are computationally demanding with limited generalization abilities for ultra-dense HetNet applications, motivating the usage of machine learning techniques for adaptive cell switching. Graph neural networks (GNNs) are powerful deep learning models with strong generalization abilities but receive little attention for cell switching. This paper proposes a GNN-based cell switching solution (GBCSS) that has a smaller computational complexity than existing heuristic algorithms. The presented performance evaluation uses the Milan telecommunication dataset based on real-world call detail records, comparing GBCSS with a traditional exhaustive search (ES) algorithm, a state-of-the-art learning-based algorithm, and the baseline without cell switching. Results indicate that GBCSS achieves a 10.41% energy efficiency gain when compared with the baseline and achieves 75.76% of the optimal performance obtained with ES algorithm. The results also demonstrate GBCSS’ significant scalability and generalization abilities to differing load conditions and the number of BSs, suggesting this approach is well-suited to ultra-dense HetNet deployment.