Abstract Integrated sensing and communication (ISAC) is a key enabler for 6G networks, yet its performance is critically constrained by the trade-off between hardware complexity and system flexibility. Conventional hybrid beamforming (HBF) architectures suffer from rigid antenna connections, limiting their adaptability in dynamic environments such as unmanned aerial vehicle (UAV) networks. To overcome this, we propose a dynamic antenna selection (AS) frame-work for HBF-based ISAC systems with joint transceiver optimization. The proposed architecture adaptively recon-figures both transmit and receive antenna subsets, forming a closed-loop system where high-precision sensing continuously guides robust communication. A weighted objective function is formulated to maximize both spectral efficiency and sensing accuracy (characterized by the Cram´er–Rao bound). To solve this mixed-integer non-convex problem, we introduce a Gumbel-Softmax relaxation that transforms the combinatorial AS task into a differentiable form, enabling gradient-based end-to-end optimization within an alternating optimization framework. Furthermore, an iterative adaptive approach (IAA) is employed for robust single-snapshot angle estimation under the dynamically formed non-uniform receive array. Simulation results demonstrate that the proposed joint transceiver optimization significantly outperforms both transmitter-only and full-array baselines, achieving a superior communication-sensing trade-off and establishing a new Pareto frontier for AS-aided ISAC systems.
Abstract Energy and material transfer and physicochemical reactions in complex industrial processes exhibit spatiotemporal propagation characteristics, and spatiotemporal relationships among variables can reveal operating patterns and improve operating performance assessment. However, numerous redundant spatiotemporal relationships interfere with operating performance representation, reducing model interpretability and cross-distribution performance stability. To address this issue, this paper proposes an operating performance assessment method for complex industrial processes based on multi-block sparse spatiotemporal graph transformer. Specifically, a perturbation response mechanism based on a conditional variational autoencoder is designed to measure the sensitivity of edge relationships to local representation changes and select perturbation-responsive edges that carry critical operating performance discriminative information. Meanwhile, a masked decay matrix is designed to impose temporal decay constraints on spatiotemporal information propagation, reflecting the continuous evolution of industrial process variables. Finally, quantile thresholds are constructed based on the sparse dynamic graph obtained from the optimal operating performance level, and key non-optimal variables are traced by analyzing edge-intensity variations. Experimental results on a coal preparation process demonstrate that the proposed method can effectively prune weakly responsive redundant relationships, achieving a pruning rate of 93.42%, while maintaining high operating performance assessment accuracy and performance stability under cross-distribution scenarios.
Abstract The Cloud-Edge-End (CEE) collaborative architecture has become a key paradigm for deploying highly intelligent multi-agent systems with low latency and high energy efficiency. Existing studies on efficient task offloading in CEE, however, either overlook long-term financial budget constraints from commercial API token usage and edge computing leasing, or inadequately handle layered information-sharing and privacy restrictions in decentralized autonomous agents. This paper investigates online inference routing to minimize the computational and communication overhead of large language model (LLM) workloads under long-term budget constraints, agent intent-sharing limitations, and un-known future prompt lengths. We propose a budget-aware online optimization framework that uses Lyapunov optimization to decouple long-term financial constraints, and show that the decoupled problem forms a generalized ordinal potential game. To address heterogeneous information-sharing restrictions, we design an adaptive best response algorithm for cooperative environments with shared decision information, and a multi-agent distributed routing algorithm for limited-information environments, enabling efficient offloading without inter-agent intent sharing. Theoretical analysis provides conditional performance and queue-length bounds and establishes finite convergence of the per-slot solvers. Simulations on distributed inference testbeds show that our scheme reduces the inference overhead of heterogeneous agentic tasks while achieving a favorable trade-off between performance and actual token expenditure.
Abstract This paper investigates an uncertain Moving-Target Traveling Salesman Problem (MT-TSP), in which the future motion trajectories of targets are uncertain. The objective is to maximize the number of visited targets within a bounded region while minimizing the total travel distance. To address the challenges introduced by dynamic target maneuvers, we propose a two-level online solution framework, termed PaE (Plan and Execute), which leverages the current observations of targets to solve the problem in real time. The high-level module plans the next target to visit, while the low-level module executes the corresponding access process. We develop a deep reinforcement learning–based algorithm at the high level, which employs an attention-based model to effectively exploit both the current environment state and historical visiting information when selecting the next tar-get. At the low level, rapid target access is achieved through a lightweight prediction mechanism with continuous state updates. Extensive experiments in simulated environments demonstrate that PaE is capable of extracting key problem features to capture dynamic changes, leveraging rich contextual information to guide decision-making, and generating high-quality solutions within extremely short computation times.
Abstract We present a novel subspace segment tree data structure designed for online aggregation queries on streaming data, addressing the challenge of efficient query processing under limited storage space in embedded systems. Traditional segment tree structures require several times the storage space of the original data, making them impractical for resource-constrained embedded devices. The proposed Subspace Segment Tree optimizes space utilization by partitioning the tree structure into three hierarchical layers, namely the solid, dynamic, and imaginary layers, and implementing a dynamic memory allocation and reclamation strategy based on a memory pool. This approach allocates memory only when nodes are actively used, significantly reducing space overhead. Experimental results demonstrate that the proposed method achieves over 90% space optimization compared to traditional segment trees while maintaining comparable time complexity, effectively resolving the space efficiency issue for aggregation queries in embedded systems.
