The rapid expansion of reusable skill repositories makes skill routing a critical capability for large language model (LLM) agents. Existing methods treat routing as task-only semantic matching. However, when users with incompatible constraints issue an identical request, this assumption conflates task relevance with skill suitability: a task-only router can select a semantically plausible skill that is unsuitable for the requesting user. To expose this failure mode, we formulate personalized skill routing as profile-conditioned retrieval, in which relevance depends jointly on the task and the user profile. We first introduce a profile-counterfactual benchmark, in which the task is held fixed while changes in the user profile induce changes in the reference skill. We further construct paired counterfactual supervision and propose SkillFeed, a progressive retrieve-and-rerank framework that first establishes task–skill alignment and then learns profile-conditioned discrimination. By retrieving body-level evidence and reranking semantically similar but profile-conflicting candidates, SkillFeed identifies skills that satisfy both task requirements and user constraints. On SkillFeed-Bench, SkillFeed attains 75.1% top-1 retrieval accuracy, a 23.1-point improvement over the corresponding pretrained routing baseline. Adding profile conditioning yields a 35.1-point gain on queries where user profile changes the reference skill. This contrast shows that user profiles are most consequential precisely when they change skill suitability. Our website is publicly available at http://www.aiskillfeed.com .

Highlights What are the main findings? A firefly-inspired clustering method and a DQN-GCN-based DGMR strategy are jointly developed to maintain resilient UAV cluster topologies under harsh network disruptions. A unified UCMR framework alleviates routing bottlenecks by exploiting neighbor edge/node features and the forwarding capacity of ordinary cluster members, improving end-to-end delay and packet delivery ratio over benchmark methods. What are the implications of the main findings? The proposed framework supports more reliable and adaptive UAV swarm communications in harsh environments with node failures, interference, and congestion. The results show the value of jointly optimizing cluster maintenance and routing, offering a practical design direction for intelligent multi-UAV networking in mission-critical scenarios.Highlights What are the main findings? A firefly-inspired clustering method and a DQN-GCN-based DGMR strategy are jointly developed to maintain resilient UAV cluster topologies under harsh network disruptions. A unified UCMR framework alleviates routing bottlenecks by exploiting neighbor edge/node features and the forwarding capacity of ordinary cluster members, improving end-to-end delay and packet delivery ratio over benchmark methods. What are the implications of the main findings? The proposed framework supports more reliable and adaptive UAV swarm communications in harsh environments with node failures, interference, and congestion. The results show the value of jointly optimizing cluster maintenance and routing, offering a practical design direction for intelligent multi-UAV networking in mission-critical scenarios.Abstract 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.

In robotic manipulation, the tight coupling between grasping and motion planning often obscures the true source of failure, leading to inefficient trial-and-error. To enable efficient long-horizon manipulation, we propose GTP-FA (Grasp-Then-Plan with Failure Attribution), a task-oriented two-stage grasp-then-plan framework that generates grasp candidates and performs downstream motion planning conditioned on the selected grasp. Given a failed manipulation trajectory, we learn a failure attribution model that generalizes to unseen grasps and produces a stable distribution over failure modes for diagnosis-guided optimization. Based on these attribution results, we then optimize both modules in a diagnosis-driven manner: on the grasping side, we inject task-level priors and risk penalties into grasp candidate scoring and optimization to suppress unstable or task-incompatible grasps; on the planning side, we target high-risk initial states through data collection and fine-tuning to address genuine planning bottlenecks. We evaluate the proposed framework in both simulation and real-robot experiments, and show that GTP-FA improves the corresponding base learners across RL, IL, diffusion-policy, and VLA-based settings, achieving substantially higher overall task success rates.

In edge-intelligent systems, efficient resource management and task scheduling are critical but challenging due to the dynamic and heterogeneous nature of edge nodes (e.g., IoT devices, drones). We model this dynamic resource allocation challenge as an online sleeping Restless Multi-Armed Bandits (RMAB) problem, where each edge node (arm) operates as a Markov decision process. Unlike prior RMAB frameworks assuming perpetual availability, our setting captures the stochastic availability of edge nodes across rounds. The system controller (learner) is unaware of the transition functions, reward distributions, and node availability a priori. The goal is to maximize expected cumulative rewards through adaptive node selection. To explore this target problem, we first derive an asymptotically optimal sleeping-index policy (SIP) as the oracle based on the fluid process transformation. Then we propose OSILA (Online Sleeping Index-aware Learning Algorithm), featuring a Minimum Exploration Guarantee (MEG) mechanism for efficient exploration. This is coupled with a modified Linear Programming-based exploitation mechanism to construct an online sleeping index, effectively handling dynamic node availability. To the best of our knowledge, this work is the first to provide the theoretical analysis (which achieves O(KT2/3logT) regret where K is the number of arms and T is the time horizon) to the online sleeping RMAB problem. Empirical results validate both theoretical guarantees and practical effectiveness in dynamic edge computing environments.

This work studies the challenge of scheduling real-time control commands in Networked Control Systems (NCS), where control actions rely on the freshness of data collected from multiple sources. In dynamic environments, ensuring that control commands in an NCS are accurate and frequent is essential for maintaining the system responsiveness. For this aim, we introduce a new metric, Age of Result (AoR), which quantifies the time elapsed since the last control command was generated and executed. This metric reflects the system’s capability to adapt to real-time changes in the operational environment by considering both data freshness and control command frequency. We conduct a detailed analysis of AoR in NCS, paying special attention to the dependencies between sensing and computing phases. We first address computation-intensive and network-intensive scenarios, proposing random sampling (RS)-based approximate algorithms for each case. Subsequently, we develop another RS-based algorithm and a heuristic approach for the general model. Simulation results demonstrate that our approach can effectively minimize AoR and significantly enhance the system performance and real-time adaptability compared to existing strategies.

Unmanned aerial vehicles (UAVs) play a critical role in disaster response, swiftly gathering information from various points-of-interest (PoIs) across extensive areas. The freshness of this information is measured by the age of information (AoI), representing the time since the latest information acquisition of a specific PoI. However, devising AoI-minimizing routes for UAVs in obstructed post-disaster environments poses unique challenges that have yet to be fully overcome. Obstacles, like post-disaster barriers, can impede direct flight paths between PoIs, and limited battery life requires energy-conscious route planning. Additionally, existing solutions fail to universally minimize varying data freshness requirements. This research addresses the AoI-driven UAV travel problem, seeking to establish periodic routes that optimize AoI metrics while considering energy and general graph constraints. We develop a learning-based algorithm to enhance the current route iteratively, utilizing guidance from a deep reinforcement learning (DRL) agent and executing a series of operations to potentially decrease AoI while adhering to topological and energy constraints. The algorithm is validated on real post-disaster datasets, demonstrating significant improvements in various AoI metrics compared to other learning-based approaches. Furthermore, our algorithm outperforms approximation algorithms and can approach the global optimum when tailored to existing AoI-minimizing problems.

Real-time object detection takes an essential part in the decision-making process of numerous real-world applications, including collision avoidance and path planning in autonomous driving systems. This paper presents a novel real-time streaming perception method named CorrDiff, designed to tackle the challenge of delays in real-time detection systems. The main contribution of CorrDiff lies in its adaptive delay-aware detector, which is able to utilize runtime-estimated temporal cues to predict objects' locations for multiple future frames, and selectively produce predictions that matches real-world time, effectively compensating for any communication and computational delays. The proposed model outperforms current state-of-the-art methods by leveraging motion estimation and feature enhancement, both for 1) single-frame detection for the current frame or the next frame, in terms of the metric mAP, and 2) the prediction for (multiple) future frame(s), in terms of the metric sAP (The sAP metric is to evaluate object detection algorithms in streaming scenarios, factoring in both latency and accuracy). It demonstrates robust performance across a range of devices, from powerful Tesla V100 to modest RTX 2080Ti, achieving the highest level of perceptual accuracy on all platforms. Unlike most state-of-the-art methods that struggle to complete computation within a single frame on less powerful devices, CorrDiff meets the stringent real-time processing requirements on all kinds of devices. The experimental results emphasize the system's adaptability and its potential to significantly improve the safety and reliability for many real-world systems, such as autonomous driving. Our code is completely open-sourced and is available at https://anonymous.4open.science/r/CorrDiff.

Wireless power transfer (WPT) is increasingly used to sustain Internet-of-Things (IoT) systems by wirelessly charging embedded devices. Mobile chargers further enhance scalability in wireless-powered IoT (WP-IoT) networks, but pose new challenges due to dynamic channel conditions and limited energy budgets. Most existing works overlook such dynamics or ignore real-time constraints on charging schedules. This paper presents a bandit-based charging framework for WP-IoT systems using mobile chargers with practical beamforming capabilities and real-time charging constraints. We explicitly consider time-varying channel state information (CSI) and impose a strict charging deadline in each round, which reflects the hard real-time constraint from the charger's limited battery capacity. We formulate a temporal-spatial charging policy that jointly determines the charging locations, durations, and beamforming configurations. Area discretization enables polynomial-time enumeration with constant approximation bounds. We then propose two online bandit algorithms for both stationary and non-stationary unknown channel state scenarios with bounded regrets. Our extensive experimental results validate that the proposed algorithms can rapidly approach the theoretical upper bound while effectively tracking the dynamic channel states for adaptive adjustment.

Federated learning (FL) is a promising approach that allows requesters (e.g., servers) to obtain local training models from workers (e.g., clients). Since workers are typically unwilling to provide training services/models freely and voluntarily, many incentive mechanisms in FL are designed to incentivize participation by offering monetary rewards from requesters. However, existing studies neglect two crucial aspects of real-world FL scenarios. First, workers can possess inherent incompatibility characteristics (e.g., communication channels and data sources), which can lead to degradation of FL efficiency (e.g., low communication efficiency and poor model generalization). Second, the requesters are budgeted, which limits the amount of workers they can hire for their tasks. In this paper, we investigate the scenario in FL where multiple budgeted requesters seek training services from incompatible workers with private training costs. We consider two settings: the cooperative budget setting where requesters cooperate to pool their budgets to improve their overall utility and the non-cooperative budget setting where each requester optimizes their utility within their own budgets. To address efficiency degradation caused by worker incompatibility, we develop novel compatibility-aware incentive mechanisms, CARE-CO and CARE-NO, for both settings to elicit true private costs and determine workers to hire for requesters and their rewards while satisfying requester budget constraints. Our mechanisms guarantee individual rationality, truthfulness, budget feasibility, and approximation performance. We conduct extensive experiments using real-world datasets to show that the proposed mechanisms significantly outperform existing baselines.

Real-time object detection is critical for the decision-making process for many real-world applications, such as collision avoidance and path planning in autonomous driving. This work presents an innovative real-time streaming perception method, Transtreaming, which addresses the challenge of real-time object detection with dynamic computational delays. The core innovation of Transtreaming lies in its adaptive delay-aware transformer, which can concurrently predict multiple future frames and select the output that best matches the real-world present time, compensating for any system-induced computational delays. The proposed model outperforms existing state-of-the-art methods, even in single-frame detection scenarios, by leveraging a transformer-based methodology. It demonstrates robust performance across a range of devices, from powerful V100 to modest 2080Ti, achieving the highest level of perceptual accuracy on all platforms. Unlike most state-of-the-art methods that struggle to complete computation within a single frame on less powerful devices, Transtreaming meets the stringent real-time processing requirements on all kinds of devices. The experimental results emphasize the system's adaptability and its potential to significantly improve the safety and reliability of many real-world systems, such as autonomous driving.

With the rapid development of edge-intelligence systems, multi-UAV platforms have become vital for large-scale 3D reconstruction. However, efficient task scheduling remains a critical challenge due to constraints on UAV energy, communication range, and the need for balanced workload distribution. To address these issues, this paper presents a novel, centralized two-stage task scheduling framework. In the first stage, the framework partitions the target area into communication-feasible subregions by applying cell decomposition that accounts for no-fly zones and workload. It then models the subregion allocation as a Capacitated Vehicle Routing Problem (CVRP) with an added balancing constraint to optimize the traversal sequence for each operational sortie. In the second stage, a time-efficient, scan-based heuristic algorithm allocates viewpoints among UAVs to ensure workload balance, minimizing the mission completion time. Extensive simulations demonstrate that our proposed approach achieves superior performance in workload balance, path efficiency, and reconstruction quality. Overall, this work provides a scalable and energy-aware solution for centralized multi-UAV 3D reconstruction, highlighting an effective approach to ensure cooperation and efficiency in complex multi-agent systems.

Mobile devices have been increasingly deployed in large-scale cyber-physical systems (CPS) to traverse the field and retrieve various data measurements from designated physical entities with stringent performance requirements. This work studies the Availability-constrained real-time Fresh Data Retrieval problem in CPS with a Speed Adjustable mobile device (AFDR-SA). The goal is to maintain the temporal validity of the real-time data with different priorities to be retrieved in the system while meeting the data availability constraints imposed by the communication range between the mobile device and the physical entities. The general case of the AFDR-SA problem is proved to be NP-hard. A dynamic programming (DP)-based optimal algorithm is proposed for a special scenario where the retrieval times of individual data items with the same priority are of the same length. For the general case where data items can have arbitrary retrieval times and different priorities, another different DP-based scheme is proposed, which is proved to be optimal given the retrieval order. A fast heuristic with low complexity is also proposed for the general problem to improve the computational efficiency. The experimental results show that the proposed schemes for the general case outperform the state-of-the-art methods and have close performance compared to the optimal solution while incurring much less computational overhead.

Unmanned Aerial Vehicle (UAV) swarm offers extended coverage and is a vital solution for many applications. A key issue in UAV swarm control is to cover all targets while maintaining connectivity among UAVs, referred to as a multi-target coverage problem. With existing dynamic routing protocols, the flying ad hoc network suffers outdated and incorrect route information due to frequent topology changes. This might lead to failures of time-critical tasks. One mitigation solution is to keep the physical topology unchanged, thus maintaining a fixed communication topology and enabling static routing. However, keeping physical topology unchanged may sacrifice the coverage. In this article, we propose to maintain a fixed communication topology among UAVs, which allows certain changes in physical topology, so that to maximize the coverage. We develop a distributed motion planning algorithm for the online multi-target coverage problem with the constraint of keeping communication topology intact. As the communication topology needs to be timely updated when UAVs leave or arrive at the swarm, we further design a topology-management protocol. Experimental results from the ns-3 simulator show that under our algorithms, UAV swarms of different sizes achieve significantly improved delay and loss ratio, efficient coverage, and rapid topology update.

We study single machine scheduling problems where the cost of processing jobs may vary over time. We assume that jobs have hard release time and deadline constraints, and each job is associated with a weight. Our goal is to select a subset of jobs to be scheduled such that (1) the total weight is maximized, and (2) a given cost budget is not exceeded. We mainly consider the case where jobs have identical processing times. We prove that this problem is NP-hard but admits a fully polynomial time approximation scheme, settling the complexity of this problem. Our results also imply that if jobs have identical weights, this problem becomes polynomial-time solvable. Then, we design more efficient algorithms for agreeable instances. Lastly, we analyze the performance of a natural greedy algorithm and show that it is strictly $3$-approximate if jobs have identical weights.

In many IoT applications, information needs to be gathered from multiple heterogeneous sources to the base station for real-time processing and follow-up actions. Undoubtedly, information freshness, measured by age of information (AoI), is critical in taking responsive actions. Recent studies have taken AoI into the consideration of transmission scheduling over wireless channels. However, existing studies on guaranteeing AoI either assume error-free wireless channels or priorly known link reliability, which is unrealistic. In this article, we tackle the AoI-guaranteed transmission scheduling problem over an unreliable channel with the aim of throughput maximization, which is modelled as an AoI-Guaranteed Multi-Armed Bandit (AG-MAB) problem. Since the problem has not been studied in the literature even for the oracle case with given link reliability, we first propose an optimal stationary randomized sampling (SRS) policy for the oracle case. For the AG-MAB problem with unknown link reliability, we propose learning algorithms that meet the AoI requirements with probability 1 and incur sublinear regret compared to Oracle SRS, which can also detect the unsatisfiability of the AoI constraint and switch to the fallback policy promptly with guaranteed accuracy. Numerical results show that our algorithm outperforms the AoI-constraint-aware baselines on throughput with per-source AoI requirement guaranteed.

Introducing mobile edge devices in IoT systems for real-time data retrieval can reduce energy consumption and human interaction, and thus has attracted significant research attentions in recent years. In the meanwhile, leveraging mobile edge device(s) to charge the sensor devices in proximity through RF-based Wireless Energy Transfer (WET) technologies can provide controllable and stable energy supply, and is being increasingly deployed in the field. Based on these two recent trends, this work aims to formulate and solve the joint device charging and fresh data retrieval problem where a speed-adjustable mobile edge device judiciously powers a set of sensor devices in emergent energy shortage through WET while retrieving designated data items through multi-hop transmissions. To this end, an efficient 3-phase method for Joint Charging and Retrieving (3JCR) is proposed. 3JCR first identifies the sensor devices to be charged and develops the moving trajectory of the mobile edge device. It then applies a max-flow min-cost based clustering scheme to determine the routing path of the required data items to be transmitted to the mobile edge device, and finally constructs the data retrieval schedule and adjusts the speed of the mobile edge device in the run time. Our extensive experimental results show that 3JCR outperforms all state-of-the-art methods in terms of the number of feasibly retrieved data items, while achieving similar node survival rate.

This work studies the Age-of-Information (AoI) optimization problem in the information-gathering wireless network systems, where time-sensitive data updates are collected from multiple information sources, and each source is equipped with a battery and harvests energy from ambient energy, such as solar, wind, etc. The arrival of the harvested energy can be modeled as the stochastic process, and an information source can deliver its data update only when 1) there is energy in the battery, and 2) this source is selected to transmit its data update based on the transmission policy. This work analyzes how the energy arrival pattern of each source and the transmission policy jointly influence the average AoI among multiple sources. To the best of our knowledge, this is the first work that formally develops the closed-form expression of average AoI in the Stationary Randomized Sampling (SRS) policy space and proposes approximation schemes with constant ratios in multi-source systems under a stochastic energy harvesting model. More specifically, under the perfect wireless channel, the closed-form expression of AoI under the SRS policy space with arbitrary finite battery size is developed. Based on the result, we propose the Max Energy-Aware Weight (MEAW) policy, which is proven to achieve 2-approximation in the full policy space. Under the uncertain wireless channel, we develop the closed-form expression of Whittle’s index to address the target problem. Based on the result, we propose the Energy-aware Whittle’s index policy (EWIP) and prove its approximate performance by using the Lyapunov optimization techniques. Experimental results show that MEAW under the perfect channel setting and EWIP under the uncertain channel setting both perform close to the theoretical lower bound and outperform the state-of-the-art schemes.

Metaverse has been viewed as the next generation of human-computer interaction, which requires collecting information from both the physical and virtual world. One potential way is to employ virtual service providers (VSPs) to finish collection tasks by designing posted-pricing mechanisms via the crowdsensing platform. As VSPs' costs and values are usually unknown, learning the optimal posted-pricing policy under uncertainty is undoubtedly critical to utilize the budget efficiently. However, existing posted-pricing learning algorithms assume that agents provide services without blocking and agents' attributes follow an independent identical distribution, both of which are unrealistic in Metaverse, e.g., VSPs should continuously sense the physical world to make provided services realistic, which makes the long working VSP unavailable/blocked for a certain period of time. In this paper, we address the budgeted pricing problem under uncertainty by considering blocking constraints and unknown non-identical VSPs' attributes. The problem is modeled as a Budgeted-pricing Blocking Bandit (B-2-bandit) problem, which remains unaddressed even for the oracle case with known VSPs' information. We thus first propose a pricing policy for the oracle case with an instance-dependent approximation ratio to the global optimum. For the general B-2-bandit problem with unknown information, we propose an online learning algorithm satisfying blocking constraints and incurring an accumulated regret up to O(MK log B) as compared to the oracle approximation algorithm, where M,K,B are the number of VSPs, candidate prices and the budget, respectively. Experiments on real datasets validate that the proposed algorithm improves more than 172% accumulated value compared to baseline pricing algorithms.

Unmanned Aerial Vehicles (UAVs) are widely used for urban image capture and large-scale 3D model reconstruction due to their cost-effectiveness and versatility. However, multi-UAV operations are limited by communication and battery constraints when covering large-scale areas. To mitigate these challenges, a common approach is to divide the target area into sub-regions and use multiple UAVs to process each sub-region sequentially. Nevertheless, achieving an equitable distribution of tasks among UAVs within these sub-regions is highly complex. To address this, we propose a well-designed heuristic-based algorithm that optimizes the allocation of predefined viewpoints across UAVs, ensuring efficient traversal and balanced workload distribution. Extensive experiments demonstrate the effectiveness of our proposed algorithm in achieving efficient and equitable task distribution.

In a crowdsensing platform, users are invited to provide data services, and multiple requesters compete for desired services. Due to users’ costs of providing services, it is critical to design incentive mechanisms to incentivize users with (monetary) rewards. Meanwhile, requesters may have individual budgets and compete for services with different procurement abilities. Such a setting falls into the budget-feasible mechanism design. However, most of the existing budget-feasible mechanisms focus on one-sided markets with a single requester rather than the two-sided markets with multiple requesters having different procurement abilities. Moreover, requesters and users can be selfish and strategic with their private information, which requires preventing information manipulation on both requesters’ and users’ sides. In this paper, we investigate budget-feasible mechanisms in two-sided crowdsensing markets where multiple strategic requesters come with private budgets to obtain services from the strategic users. We also consider the fairness on the requesters’ side, i.e., a requester with more budget should obtain more service. We propose budget-feasible mechanisms for two models by distinguishing the types of services, i.e., the homogeneous or heterogeneous services. All proposed mechanisms satisfy fairness, budget feasibility, truthfulness on both users’ and requesters’ sides, and the constant approximation ratio. Numerical experiment results further demonstrate the efficiency of our proposed mechanisms.