Time-sensitive networking (TSN) provides deterministic communication for industrial applications, but mixed-criticality scheduling remains a difficult combinatorial problem. Large language models (LLMs) offer flexible in-context search, yet their proposals may be incomplete or violate scheduling constraints. We present CLEAR-TSN (Constraint-Checked LLM Evolution with Adaptive Repair for TSN), an LLM-guided evolutionary-search framework for traffic-class assignment over predefined routes. It combines population-based exploration, adaptive temperature, constrained repair and local replenishment with deterministic scheduling and score feedback. The LLM proposes assignments, while local validation and controller approval govern deployment. Three-seed comparisons across five SAE loads and four additional network configurations show the highest mean nominal-utilization score in seven of nine settings.
Low-altitude uncrewed aerial vehicle (UAV) networks are becoming essential infrastructure underpinning aerial services, including emergency response, aerial inspection, and urban management. Characterized by highly dynamic three-dimensional topologies and heterogeneous node resources, flying ad hoc networks (FANETs) face substantial challenges in maintaining reliable routing and executing computational tasks efficiently. This article proposes a node state sensing-based dual-layer predefined-time convergence framework (NSS-DPCF) for low-altitude FANETs. By incorporating node operational states into routing and task offloading, NSS-DPCF enables adaptive offloading among idle, partially available, and fully utilized UAV nodes. The proposed control architecture couples global resource orchestration and routing optimization at the optimization layer with rapid local recovery at the tracking layer, thereby enabling the network to respond to dynamic topology variations within a predefined-time bound. Quantitative evaluations demonstrate that, relative to DRL-FIR, NSS-DPCF decreases end-to-end delay and routing overhead by 25.1% and 22.2%, respectively. Furthermore, this article discusses promising research directions, including mobility-aware predictive routing, trustworthy node-state perception, and integrated communication-computation-control optimization. Finally, this work provides a resource-adaptive routing and orchestration paradigm for future intelligent aerial networks.
With the rapid growth of the low-altitude economy, accurate environmental sensing is essential for drone safety. While multi-drone systems overcome the blind spots of single agents by sharing views, real-world wireless networks are severely hindered by communication barriers and malicious node attacks. In this paper, we propose the robust framework termed Fault-Tolerant Collaborative Perception (FTCP) to overcome these inherent vulnerabilities. Initially, we construct a hierarchical system architecture spanning from raw data acquisition to object optimization, systematically analyzing the communication barriers and security challenges encountered throughout this process. Subsequently, we conduct an exploration of the key technologies integrated within the data transmission and data fusion layers, introducing latency-aware compensation, critical data transmission, packet drop recovery, and distributed anomaly detection. To validate our approach, we present CoSwarm, a novel multi-modal dataset tailored for secure low-altitude perception. The experimental results show that FTCP improves detection accuracy and system resilience under communication impairments and malicious-node interference, supporting reliable low-altitude perception for smart-city applications.
Low-altitude wireless networks (LAWNs) are expected to support highly mobile aerial users over rapidly varying air-to-ground links with pronounced delay-Doppler selectivity. These features make orthogonal time frequency space (OTFS) modulation attractive, but they also impose stringent requirements on detector reliability and computational efficiency. A typical detection issue is that a low-complexity detector may produce unreliable estimates at certain DD-domain positions. Based on this observation, this paper presents a practical global-to-local two-stage detection framework for OTFS in LAWNs. The key idea is to employ a two-stage architecture that combines a low-complexity global soft-output detector with selective local refinement. The first stage processes the entire DD grid to obtain initial soft information and symbol estimates. The second stage then constructs local observation equations for unreliable positions, generates candidate vectors based on test error patterns (TEPs), and computes refined symbol-level soft information. The case study demonstrates that the second-stage refinement consistently improves the single-stage detector under both uncoded and coded transmission. It also illustrates the impact of the number of local candidate searches on detection performance and the relationship between Stage-II computational complexity and SNR. These results indicate that reliability-driven local refinement is a promising approach to practical OTFS detection in future low-altitude networking scenarios.
Low-altitude wireless networks (LAWNs) are emerging as a key enabling infrastructure for the low-altitude economy (LAE). Integrated sensing and communication (ISAC) has been widely advocated to improve spectrum efficiency and situational awareness in such networks. However, in low-altitude environments where aerial transmissions may be subject to regulatory monitoring or adversarial surveillance by a warden, the integration of sensing fundamentally alters signal observability, raising a critical question: Can ISAC operate covertly under such conditions? This article rethinks ISAC from an observability perspective, with a focus on low-altitude networks characterized by dense sensing infrastructures and strong line-of-sight exposure. Our central argument is that ISAC can approach covert operation in LAWNs when covertness is explicitly treated as a system-level design constraint, rather than as a byproduct of communication or sensing-oriented optimization. By examining sensing-induced observability and representative warden models, we discuss when and under what conditions ISAC-enabled systems can approach covert operation. Our analysis reveals inherent tradeoffs between sensing, communication, and covertness, and highlights key challenges and design considerations for stealth-aware ISAC in future low-altitude wireless networks.
