This article presents findings from a multimodal autonomous delivery pilot project conducted in Arlington, TX, with active community participation. Over one week, an integrated system combining an uncrewed aircraft system (UAS) and an autonomous ground vehicle (AGV) completed 139 scheduled food box deliveries to residents. This article highlights the operational potentials and challenges revealed through the demonstration and introduces mathematical models to evaluate four key factors-risk, cost, environmental impact, and delivery time-that affect the delivery process. Recognizing that stakeholders prioritize these factors differently, the article proposes a guideline for formulating delivery planning and scheduling as a mixed optimization problem.

High specularity, weak texture, and noisy pseudo-depth cues make capsule defect detection particularly challenging in practice. Existing RGB-D detectors often rely on direct fusion strategies that are sensitive to depth noise, while physical depth sensors remain unstable or costly under reflective conditions. This paper studies a stage-wise RGB-D-Text detection framework for capsule defect detection under noisy pseudo-depth inputs. Rather than introducing a fundamentally new detector, the proposed framework adapts a strong vision-language detector to this setting through three task-oriented components: (1) a Learnable Depth Calibration (LDC) module that separates low-frequency structure from high-frequency residuals for lightweight depth correction; (2) a Modality-Aware Selective Fusion (MSF) mechanism that progressively integrates RGB and depth features across pyramid levels; and (3) an Anchor-Guided Cross-Modal Attention (AGCA) module that injects a small set of fixed category-level text prompts into compact visual anchors for semantic refinement. On the Capsule-Defect-RGBD dataset, this staged design achieves 85.6

Decision-making for systems with high-dimensional uncertain parameters is critical to maintaining system stability and reliable operation. This paper develops an efficient decision-making method for high-dimensional uncertain systems by leveraging the Multivariate Probabilistic Collocation Method (MPCM)-Taguchi method. Based on the MPCM-Taguchi framework, an integral reinforcement learning (IRL) control algorithm is developed. The proposed method is applied to power buffer systems in DC microgrids with uncertain loads. Comparative studies demonstrate the effectiveness and practical value of the approach for real-time decision-making in microgrid systems.

The evaluation of output statistics in systems with high-dimensional uncertain parameters is important for real-time decision-making in large-scale systems. In this paper, we develop an MPCM-Taguchi method that combines the Multivariate Probabilistic Collocation Method (MPCM) with the Taguchi method to accurately estimate system output statistics while significantly reducing the number of simulations. The estimation algorithm is developed, and its theoretical analysis is provided. Numerical studies and the application to power buffers in Direct Current (DC) Microgrids with uncertain loads validate the MPCM-Taguchi uncertainty evaluation method.

CLIP-based video-text retrieval models are efficient, but their video branch often relies on a small set of sampled frames and a simple aggregation rule. This is sufficient for many object-centered queries, yet it weakens the description of action order and cross-frame dependencies. We study this issue under a frozen CLIP4Clip-style setting and add a lightweight module, named STAA, after the CLIP visual encoder. STAA introduces relative temporal encoding, a graph-guided attention bias over local, global, and motion-related frame relations, and two refinement streams for temporal and channel-level adjustment. The backbone is kept unchanged, so the added module only modifies the frame-token representation before retrieval matching. Experiments on MSR-VTT, MSVD, and ActivityNet Captions show consistent improvements over the reproduced CLIP4Clip baseline. On MSR-VTT text-to-video retrieval, R@1 increases from 44.5 to 46.8, and repeated runs give 46.7 ± 0.2 compared with 44.5 ± 0.3 for the baseline. The module adds 2.7

While Vision Transformers (ViT) have revolutionized global semantic modeling, their fundamental patch-based embedding strategy inevitably acts as a low-pass filter, degrading high-frequency geometric information and resulting in blurred boundaries for object detection. Existing feature pyramid networks often focus on semantic consistency but overlook the explicit recovery of these physically lost local structural cues. To address this, we propose the Edge Progressive Feature Pyramid Network (EPFPN). Unlike generic fusion schemes, EPFPN incorporates a novel Edge-Level Attention mechanism specifically designed to reconstruct directional edge features from flattened patch representations. By progressively fusing these recovered high-frequency details with deep semantics via the Edge-Pixel Fusion module, our method effectively compensates for the spatial quantization errors inherent in ViT backbones. Extensive experiments demonstrate that EPFPN consistently outperforms state-of-the-art baselines on MS COCO. Notably, it boosts the powerful Hybrid Task Cascade framework with a ViT-B backbone to 45.9

Occlusion remains a critical challenge in multi-object tracking (MOT), often causing prediction errors and identity switches when using conventional Kal-man filter-based methods. To address this issue, we propose a novel two-stage association framework incorporating a Noise-Optimized Kalman Filter (NOKF) and a confidence-scaling mechanism. Unlike prior adaptive Kalman filters that adjust only a single noise source, NOKF dynamically updates both measurement and process noise covariances using a sliding window and exponential moving average, improving motion prediction robustness under occlusion. To further reduce association errors, we introduce Confidence-Scaling Efficient Intersection over Union (CS-EIoU), which integrates detection confidence with motion cues to better handle low-confidence detections and fast-moving targets. A two-stage association strategy processes high-confidence and low-confidence detections separately, enhancing trajectory continuity and target recall. Extensive experiments on MOTChallenge benchmarks demonstrate that our method outperforms recent tracking by detection methods. Experiments on the DanceTrack dataset validate the robustness of the proposed method. Ablation studies and visualizations further confirm the effectiveness of each proposed component, especially in densely occluded scenes.

In unmanned aerial vehicle (UAV) air-to-air (A2A) communication with rigidly mounted directional antennas, flight-induced attitude variations can cause beam deflections, leading to degraded link quality or even complete signal loss. Existing studies typically assume level flight, thereby overlooking this critical effect. In this paper, we formulate communication reliability as a yaw alignment and pitch-constrained control problem, where yaw regulation ensures beam pointing and pitch is restricted within the half-power beamwidth (HPBW) to preserve link reliability. To address this formulation, we design a time-varying gain command-filtered backstepping controller that achieves accurate trajectory tracking while rigorously enforcing attitude constraints. Theoretical analysis shows that the closed-loop signals converge within a prescribed time, thus avoiding large transient responses. Simulation demonstrates that the proposed method sustains stable and reliable communication under representative flight scenarios.

In industrial scenarios, existing deep generative methods face the challenge of adapting to new domains with limited data for anomaly detection. Flow-based generative models are efficient but fail to perform well in out-of-distribution detection due to the inductive bias. To solve these problems, we utilize a hierarchical normalizing flow framework for few-shot anomaly detection and localization (FesFlow). First, we propose a multi-scale attention coupling block in each single flow with channel and spatial self-attention mechanism, which can model long-term dependencies of flows and obtain fine-grained anomalous feature distribution. Furthermore, to reduce the bias towards local-pixel correlation, we introduce a global and local aggregation module to capture semantic context and fuse with low-level detailed features. Compared to state-of-the-art methods, our approach demonstrates outstanding performance evaluated on MVTec-AD and BTAD datasets, achieving optimal balances in both detection and localization tasks in an end-to-end network.

Tiny Person detection in long-range scenes is a popular and challenging task. Current person detectors have two major issues. Firstly, their performance is poor in the case of tiny and heavily occluded persons. Secondly, they are computation-intensive and have large model sizes, which make them difficult to deploy on resource-limited devices. To solve the above issues, we proposed TPS-YOLO. Based on YOLOv8, we reconstruct the network structure by introducing shallow features of P2 into the feature fusion layers, which helps retain more spatial information important for tiny person detection. We design a fine-grained feature extraction module SPDCA to replace the standard convolution layer in the backbone network to enhance the feature representation of the network. In the feature fusion network, we use a weighted fusion method to fuse multi-scale features, which introduces learnable weights to learn the importance of different input features. We propose a lightweight module named C2f_Efficient, which integrates Depthwise Separable Convolution (DSC) to reduce the model parameters. Furthermore, we apply a model pruning method to further reduce the model's computational complexity. Experiments on the Tinypersonv2 and VisDrone-person datasets show that TPS-YOLO achieves satisfactory performance in terms of both efficiency and accuracy and has advantages on model lightweight.

Safe and efficient decision-making is crucial for autonomous vehicles (AVs) to operate in complex traffic scenarios. The uncertain intentions of surrounding agents (e.g., AVs, human-driven vehicles, and pedestrians) and their complex interactions pose significant challenges to the decision-making process. In this work, we propose a novel stochastic hierarchical game (SHG) framework to address the challenges of decision-making under uncertain intentions of agents in complex multi-vehicle settings. The hierarchical game approach addresses the scalability of multi-agent interactions. In the stochastic game framework, the uncertain intentions of traffic agents are captured using random mobility models (RMMs), the parameters of which can be estimated using data-driven methods. We propose a novel decision-making method for the ego vehicle based on the expected optimal actions under uncertain intentions. To solve the SHG efficiently, an uncertainty quantification method called the multivariate probabilistic collocation method with an orthogonal fractional factorial design (MPCM-OFFD) is deployed. Comparative simulation studies are conducted to verify the effectiveness of the proposed framework.

This paper develops a game-theoretic decision-making framework for autonomous driving in multi-agent scenarios. A novel hierarchical game-based decision framework is developed for the ego vehicle. This framework features an interaction graph, which characterizes the interaction relationships between the ego and its surrounding traffic agents (including AVs, human driven vehicles, pedestrians, and bicycles, and others), and enables the ego to smartly select a limited number of agents as its game players. Compared to the standard multi-player games, where all surrounding agents are considered as game players, the hierarchical game significantly reduces the computational complexity. In addition, compared to pairwise games, the most popular approach in the literature, the hierarchical game promises more efficient decisions for the ego (in terms of less unnecessary waiting and yielding). To further reduce the computational cost, we then propose an improved hierarchical game, which decomposes the hierarchical game into a set of sub-games. Decision safety and efficiency are analyzed in both hierarchical games. Comprehensive simulation studies are conducted to verify the effectiveness of the proposed frameworks, with an intersection-crossing scenario as a case study.

This abstract describes the project funded by the IEEE SMCS on Transforming Educational Assets and Materials (TEAM) in Systems, Man, and Cybernetics. The project develops an open course on Cyber Physical Systems (CPS) Foundation and Design for Unmanned Aerial Vehicles (UAVs). The course will be available to the public and serve the need of researchers, students and professionals who are interested in conducting UAVs related research. The open course contains integrated modules on control, communication and networking, computing, and artificial intelligence (AI) to provide trainees a comprehensive knowledge needed for UAVs. The course is self-paced and contains quizzes in each module for help students assess the quality of learning and also allow course designers to evaluate the effectiveness of the course materials for continuous improvement. The open course promotes CPS which is a SMCS technical field. It will also attract students and professionals to the SMC community.

The development of unmanned aerial systems (UASs), a.k.a., drones, is a highly cross-disciplinary subject, demanding diverse expertise to build comprehensive systems from scratch. The introduction of open and standardized hardware testbeds has the potential to accelerate UAS development and mitigate the barrier to entry in this rapidly growing field. In this paper, we introduce an open platform based on Kubernetes that offers users remote access to a shared network of hardware testbeds consisting of UASs, ground rovers and compute devices. Three use cases are presented to showcase the key features of the proposed platform and its potential to facilitate research and education in UAS, robotics and edge computing.

Networked airborne computing (NAC) has witnessed a notable surge in interest over recent years. This emerging technology promises to enhance the onboard computing capabilities of Unmanned Aerial Vehicles (UAVs), potentially revolutionizing both current and future applications. Despite its promise, there are substantial challenges impeding the realization of a high-performance NAC system. Among these, reducing communication delays between UAVs is a critical issue that needs immediate attention. Traditional sub-6GHz technologies prove inadequate in meeting the low-latency requirements essential for NAC applications that involve high data traffic. A promising solution to this challenge is the adoption of millimeter wave (mmWave) frequency bands for Air-to-Air (A2A) communications, which offer a wealth of spectrum resources. However, the integration of mmWave within NAC systems remains under-explored. To bridge this gap, this paper presents a comprehensive study on the design of advanced NAC systems, integrating mmWave for A2A communications. A holistic analysis from both hardware and software perspectives is conducted, providing valuable insights that will facilitate the development of advanced NAC systems with high airborne computing performance.

Safety and efficiency are primary goals of air traffic management. With the integration of unmanned aerial vehicles (UAVs) into the airspace, UAV traffic management (UTM) has attracted significant interest in the research community to maintain the capacity of three-dimensional (3D) airspace, provide information, and avoid collisions. We propose a new decision-making architecture for UAVs to avoid collision by formulating the problem into a multi-agent game in a 3D airspace. In the proposed game-theoretic approach, the Ego UAV plays a repeated two-player normal-form game, and the payoff functions are designed to capture both the safety and efficiency of feasible actions. An optimal decision in the form of Nash equilibrium (NE) is obtained. Simulation studies are conducted to demonstrate the performance of the proposed game-theoretic collision avoidance approach in several representative multi-UAV scenarios.

In this study, we investigate the resilience of DC microgrids in the face of disturbances that could induce boost converter failures. We associate the converter failure conditions with disturbances and implement a power buffer control system, which prevents voltage collapse and promotes system stability. A new resilience model is proposed that considers general power mismatches for a comprehensive resilience evaluation. We further evaluate the resilience of an interconnected DC microgrid where the stability of the system is ensured through proofs and examine the role of power buffer control in enhancing resilience against disturbances. The results validate the significance of power buffer control in augmenting DC microgrid resilience. The hardware-in-the-loop experiment study demonstrates over $32\%$ improvement of resilience using the proposed control.