Yan Lyu

Reliable UAV navigation is challenging when direct geometric perception is limited and only sparse environmental observations are available. This paper presents CoSafe-Nav, an integrated navigation framework that uses ultra-wideband (UWB) line-of-sight (LOS) observations to construct a candidate corridor and guide trajectory generation. The framework addresses fixed-altitude planar navigation with known, pre-deployed anchors and an available UAV pose estimate. LOS-supported segments are accumulated online, and a grid-based Euclidean distance transform (EDT) provides a common clearance representation for path ranking and trajectory refinement. The path planner combines travel distance with a local inverse-distance penalty. The trajectory stage corrects interior control points using the EDT gradient, regenerates the spatial curve, and assigns execution timing. Evaluation comprises two simulation scenarios, component comparisons, an anchor-availability study, and indoor UAV trials. CoSafe-Nav completed all ten navigation tasks in each simulation scenario and four of five indoor trials. In the path-planning comparison, mean EDT corridor clearance increased from 0.76 to 0.98 m and from 0.63 to 0.91 m, accompanied by longer routes. These means describe successful tasks within each configuration. The trajectory comparison also showed a lower acceleration integral for the safety-guided method. The results support the feasibility of the integrated processing chain under the stated pre-instrumented deployment conditions.

Realistic modeling of pedestrian flow in dense public spaces is important for urban crowd geosimulation, mobility analysis, and indoor public-space geo-information modeling. Although prior research has emphasized microscopic agent interactions, higher-level spawn dynamics—governing when and where pedestrians appear—remain less explored, despite their fundamental role in shaping crowd density and flow. Existing approaches often decouple spatial and temporal generation, limiting their ability to capture rich spatio-temporal correlations, and they lack controllability for user-specific scenarios such as high-density environments. In this paper, we propose a Guided Joint Spatio-Temporal Diffusion framework for pedestrian spawn simulation. Our objective is to develop and evaluate a controllable joint spatio-temporal generative model that produces each pedestrian spawn event—its inter-arrival time, origin, and destination—consistent with observed spawn dynamics and a user-specified normalized local spawn-intensity condition. The model addresses the upstream initialization of a crowd simulation, rather than complete trajectory prediction or microscopic interaction simulation, and is evaluated through both next-event accuracy and fixed-horizon controllability. The method leverages spatio-temporal diffusion point processes to jointly model spatial and temporal spawn events, capturing dependencies overlooked by classical and neural point-process-based methods. To support controllable pedestrian-flow generation for geosimulation and downstream applications, we integrate a conditional denoising network with classifier-free guidance, enabling user-specified factors such as crowd density to steer generation. Experiments on the Grand Central dataset demonstrate that our method outperforms strong baselines, reducing temporal error (T-RMSE) by 38% and achieving consistent improvements in spatial and spatio-temporal accuracy. These results show the potential of diffusion-based spatio-temporal modeling for controllable urban crowd geosimulation and pedestrian mobility data generation.

UAV swarms provide a flexible sensing platform for smart-city crowd surveillance, but cooperative aerial monitoring remains challenging due to dynamic pedestrian distributions, partial observability, and the trade-off between visual coverage and recognition accuracy. In particular, flying at higher altitudes increases the field of view but reduces recognition accuracy, while low-altitude flight improves visual quality at the cost of limited coverage. To address these challenges, this paper proposes an environment-aware cooperative navigation framework that integrates spatiotemporal density prediction with multi-agent reinforcement learning. The surveillance area is modeled as a spatiotemporal graph, where sparse and partial UAV observations are used to predict future pedestrian-density maps and confidence intervals. The predicted density and uncertainty, together with empirical recognition error, UAV position, flight height, battery state, and historical observations, are incorporated into MARL-based policy learning. The learned policy enables UAVs to cooperatively adjust movement and altitude decisions under the centralized training and decentralized execution paradigm. Extensive simulations in UAV-based crowd surveillance environments demonstrate that the proposed framework achieves a more favorable coverage–error trade-off than representative heuristic, prediction-based, single-agent reinforcement learning, and multi-agent reinforcement learning baselines. The results show that prediction-aware and accuracy-aware cooperation improves pedestrian-level surveillance performance under dynamic and partially observable crowd distributions.

UAV clusters operating in harsh environments must maintain connectivity under sudden node failures and dynamic traffic demands. Existing studies often address maintenance and routing separately, which leads to unnecessary reorganization overhead and forwarding bottlenecks around gateway nodes. To overcome these limitations, we propose a unified UAV Cluster Maintenance and Routing System (UCMR) that combines a firefly-inspired clustering method, a DQN–GCN-based medium-level interference recovery strategy (DGMR), and load-aware routing over forwarding neighbors. The maintenance component improves backbone repair by integrating global topology features with local observations, while the routing component exploits edge and node features to distribute forwarding pressure beyond gateway nodes. Extensive experiments demonstrate that UCMR outperforms representative benchmark methods in terms of average end-to-end delay and packet delivery ratio.

Visual simultaneous localization and mapping (SLAM) is vulnerable to drift in dynamic environments because moving objects create erroneous data associations, whereas indiscriminate removal of potentially dynamic regions depletes useful visual features. This study presents AdaDyn-SLAM, an RGB-D–inertial framework that combines hierarchical geometric–semantic feature classification, semantic-aware loop closure, and reliability-aware robust optimization. Instance masks and multi-view geometry partition observations into static, semi-static, and dynamic classes; temporal object evidence is converted into continuous visual-residual weights so that temporarily stationary objects remain usable without being treated as fully reliable. For loop closure, local semantic class distributions complement binary local-appearance descriptors. The sliding-window back end adjusts visual robustification from the strict-static-feature ratio while retaining inertial constraints when visual reliability declines. On OpenLORIS-Scene, the complete system obtains ATE RMSE values of 0.369 m on Cafe_1 and 1.594 m on Corridor_4. On VCU-RVI Lab-dynamic4, its three-replay mean ATE RMSE is 0.148 m with a sample standard deviation of 0.009 m, 60.9% below the Dynamic-VINS mean of 0.379 m. Across the five VCU-RVI dynamic sequences, its translational and rotational RPE range from 0.005–0.009 m/s and 0.339–1.568 deg/s, respectively. Three closed-route mobile-robot trials produce endpoint-return drifts of 0.080, 0.130, and 0.105 m. The gains are scene dependent: Lab-dynamic5 exhibits worse global translational ATE despite improved local RPE, exposing a limitation of the current global reliability model. On an Intel Core i7-11800H and NVIDIA RTX 3070, AdaDyn-SLAM processes 27.4 frames/s and requires 3.00 GiB of RAM and 1.45 GiB of GPU memory. These results demonstrate the benefit of motion-aware feature retention under substantial dynamic interference while quantifying its operating envelope and computational overhead.

In edge-assisted low-latency video analytics, a critical challenge is balancing on-device inference latency against the high bandwidth costs and network delays of offloading. Ineffectively managing this trade-off degrades performance and hinders critical applications like autonomous systems. Existing solutions often rely on static partitioning or greedy algorithms that optimize for a single frame. These myopic approaches adapt poorly to dynamic network and workload conditions, leading to high long-term costs and significant frame drops. This paper introduces a novel partitioning technique driven by a Deep Reinforcement Learning (DRL) agent on a local device that learns to dynamically partition a video analytics Deep Neural Network (DNN). The agent learns a farsighted policy to dynamically select the optimal DNN split point for each frame by observing the holistic system state. By optimizing for a cumulative long-term reward, our method significantly outperforms competitor methods, demonstrably reducing overall system cost and latency while nearly eliminating frame drops in our real-world testbed evaluation. The primary limitation is the initial offline training phase required by the DRL agent. Future work will focus on extending this dynamic partitioning framework to multi-device and multi-edge environments.

Personalized Federated Learning (PFL) aims to train a personalized model for each client that is tailored to its local data distribution, learning fails to perform well on individual clients due to variations in their local data distributions. Most existing PFL methods focus on personalizing the aggregated global model for each client, neglecting the fundamental aspect of federated learning: the regulation of how client models are aggregated. Additionally, almost all of them overlook the graph structure formed by clients in federated learning. In this paper, we propose a novel method, Personalized Federated Learning with Graph Attention Network (pFedGAT), which captures the latent graph structure between clients and dynamically determines the importance of other clients for each client, enabling fine-grained control over the aggregation process. We evaluate pFedGAT across multiple data distribution scenarios, comparing it with twelve state of the art methods on three datasets: Fashion MNIST, CIFAR-10, and CIFAR-100, and find that it consistently performs well.

Ride-hailing platforms have been facing the challenge of balancing demand and supply. Existing vehicle reposition techniques often treat drivers as homogeneous agents and relocate them deterministically, assuming compliance with the reposition. In this paper, we consider a more realistic and driver-centric scenario where drivers have unique cruising preferences and can decide whether to take the recommendation or not on their own. We propose i-Rebalance, a personalized vehicle reposition technique with deep reinforcement learning (DRL). i-Rebalance estimates drivers' decisions on accepting reposition recommendations through an on-field user study involving 99 real drivers. To optimize supply-demand balance and enhance preference satisfaction simultaneously, i-Rebalance has a sequential reposition strategy with dual DRL agents: Grid Agent to determine the reposition order of idle vehicles, and Vehicle Agent to provide personalized recommendations to each vehicle in the pre-defined order. This sequential learning strategy facilitates more effective policy training within a smaller action space compared to traditional joint-action methods. Evaluation of real-world trajectory data shows that i-Rebalance improves driver acceptance rate by 38.07% and total driver income by 9.97%.

Ride-hailing platforms have been facing the challenge of balancing demand and supply. Existing vehicle reposition techniques often treat drivers as homogeneous agents and relocate them deterministically, assuming compliance with the reposition. In this paper, we consider a more realistic and driver-centric scenario where drivers have unique cruising preferences and can decide whether to take the recommendation or not on their own. We propose i-Rebalance, a personalized vehicle reposition technique with deep reinforcement learning (DRL). i-Rebalance estimates drivers' decisions on accepting reposition recommendations through an on-field user study involving 99 real drivers. To optimize supply-demand balance and enhance preference satisfaction simultaneously, i-Rebalance has a sequential reposition strategy with dual DRL agents: Grid Agent to determine the reposition order of idle vehicles, and Vehicle Agent to provide personalized recommendations to each vehicle in the pre-defined order. This sequential learning strategy facilitates more effective policy training within a smaller action space compared to traditional joint-action methods. Evaluation of real-world trajectory data shows that i-Rebalance improves driver acceptance rate by 38.07% and total driver income by 9.97%.